Course: BAYESIAN LEARNING  2021 – Professional Master in Economics
Professor: Hedibert Freitas Lopes – www.hedibert.org

Teaching assistant: Igor Ferreira Batista Martins – igorfbm@al.insper.edu.br

Syllabus

Midterm exam (take-home) (solution) +

Homework assignments

1. HW1 (Solution) (Code)
2. HW2 (Code with solution)
3. HW3
4. HW4

Examples developed in class: 

• Class of 05/11/2021: Normal-Normal and Poisson-Gamma models
• Class of 05/11/2021: A bit of Monte Carlo simulation and integration
• Class of 05/11/2021: Back to the physicists example: MC simulations (results) (html)
• Class of 05/18/2021: Normal vs skew-normal distributions: an exercise in Bayesian learning
• Class of 05/18/2021: Skew-normal vs skew-t distributions + finite mixtures
• Class of 05/18/2021: Road map to study Chapters 2 and 3 of Gamerman and Lopes (2006)
• Class of 05/18/2021: SIR for a multimodeal bivariate posterior distribution (results)
• Class of 05/18/2021: SIR for the sums of heads when flipping two coins.
• Class of 05/21/2021: The potential fallacy of using the prior as proposal in SIR algorithms
• Class of 05/21/2021: Regressing weight on height: SIR for Gaussian and Student’s t models
• Class of 05/21/2021: Bayesian learning of correlation in bivariate normal data via SIR algorithm
• Class of 06/01/2021: Bayesian Poisson regression model versus i.i.d. Poisson model
• Class of 06/01/2021: Bayesian linear modeling (BLM): Posterior inference and model comparison (R code)
• Class of 06/01/2021: BLM: my first MCMC algorithm (wnba/nba data)
• Class of 06/01/2021: BLM: subset selection via BIC and shrinkage/sparsity via regularizing priors – Stock and Watson’s (2002) macro data
• Class 0f 06/01/2021:  BLM: k-fold CV (Stine and Foster’s (2014) dataset) (Rcode)
• Class of 06/01/2021: BLM: In-sample vs out-of-sample MSE/MAE (data) (R code)
• Class of 06/08/2021: The Binomial model and Beta prior: another illustrative example of the Bayesian paradigm
• Class of 06/15/2021: Deny mortgage?  Logit/probit model
• Class of 06/15/2021: Jim Albert’s example – Bayesian probit model via Gibbs sampler
• Class of 06/15/2021: Logistic model – ICU data: glm, bayesglm & glmnet (ridge, elastic net, lasso) (R code)
• Class of 06/15/2021: Receiver operating characteristics (ROC) and area under the curve (AUC)
• Class of 06/15/2021: Classifying player as WNBA or NBA player based on the player’s height (AIC & misclassification)
• Class of 06/22/2021: Revisiting the ICU data: CART, BART and random forest (R code)
• Class of 06/22/2021: Revisiting Stock and Watson’s (2002) macro data
• Class of 06/22/2021: Four BART applications & 2 reviews + cute CART trees and 3D plots
• Class of 06/29/2021: Bayesian two layer neural network  (R code) – More comparisons (R code)

Course notes (+ R code & references)

• Bayesian ingredients  
  • Bayesian update: a toy example
  • Flipping one of three coins three times and observing three heads (R code)
  • Chapter 1 of “Bayes’ Rule: A tutorial Introduction to Bayesian Analysis (Slide)
  • Tiago Mendonca’s shiny for the physicists example
• Bayesian computation
  • iid Bernoulli or logit regression or probit regression? (Rmarkdown)
  • Learning the number of degree of freedom of a Student’s t  (Rmarkdown)
  • Bayesian regression with the normal-gamma (NG) prior
  • Bayes factor (BF) and Bayesian model averaging (BMA) via sampling importance resampling (SIR)
  • Smith and Gelfand (1992) Bayesian Statistics without Tears: A Sampling-Resampling Perspective. The American Statistician, 46(2), 84-88.
• Bayesian linear regression
  • Motorcycle example
  • A few R packages for Bayesian inference in linear models
  • Boston housing data: ML and Bayesian inference
  • Stan/rstan: Michael Betancourt’s Bayes Sparse Regression

• Bayesian classification via logistic regression
  • Sparse logistic regression for the spam/ham dataset (data)
  • Marketing campaigns of a Portuguese banking institution (data)
  • Sparse logistic regression: comparison of regularization and Bayesian implementations
  • Gelman, Jakulin, Pittau and Zu (2008) A weakly informative default prior distribution for logistic and other regression models, AOAS, 2(4), 1360-1383.
  • Polson, Scott and Windle (2013) Bayesian Inference for logistic models using Pólya-Gamma latent variables, JASA, 108, 1339-1349.

• Other important modeling structures
  • Factor models (Additional material)
    • Indice de Desenvolvimento Humano
    • US market (data)
    • Bayesian factor models
    • Reducao de ruido com PCA e FA
  • Finite mixture of distributions
• Treed models
  • Classification and regression trees (CART)
    • CART step-by-step for the Boston housing data
    • CART for the motorcycle data
    • ROC and AUC
  • Bayesian CART
  • Bootstrap aggregating  (bagging)
    • Ilustracao do bootstrap
  • Boosting (weak/stronger learners)
  • Random forests
    • Predicao da covid-19 com dados publicos do Hospital Israelita Albert Einstein/SP
  • Bayesian additive regression trees (BART) + (Rob’s notes) + (a list of slides/examples/people/papers)
• Latent Dirichlet Allocation (LDA) + short list of slides and papers
  • Example 1 & 2: Twitter & BBC
  • Example 3: Dickens, Wells, Verne or Austen?
• Neural Networks + short list of slides and papers

Additional supporting material

• Stan/rstan for posterior inference: Hamiltonian MC (HMC) methods – by Hedibert Lopes
• MC and MCMC: Key References – by Hedibert Lopes
• Mathematics for Machine Learning – by Deisenroth, Faisal and Ong
• Conceitos e analises estatisticas com R e JASP – by  Luis Anunciacao
• Data Science, Marketing and Business (by Pedro Fernandes & Paulo Marques)
• Aprendizado de Máquina: Uma Abordagem Estatística (by Rafael Izbicki & Tiago Mendonça)
• Estatística e Ciência de Dados (by Pedro Morettin & Julio Singer)

Course: ADVANCED BAYESIAN ECONOMETRICS 2021 – Doctoral Program in Business Economics
Professor: Hedibert Freitas Lopes – www.hedibert.org

Objective: The end of the course goal is to allow the student to critically decide between a Bayesian, a frequentist or Bayesian-frequentist compromise when facing real world problems in the fields of micro- and macro-econometrics and finance, as well as in quantitative marketing, strategy and business administration.  With this end in mind, we will visit well known Bayesian issues, such as prior specification and model comparison and model averaging, but also study regularization via Bayesian LASSO, Spike-and-Slab and related schemes, “small n, large p” issues, Bayesian statistical learning via additive regression trees, random forests, large-scale VAR and (dynamic) factor models.

Course description: Basic ingredients: prior, posterior, and predictive distributions, sequential Bayes, conjugate analysis, exchangeability, principles of data reduction and decision theory.  Model criticism: Bayes factor, computing marginal likelihoods, Savage-Dickey ratio, reversible jump MCMC, Bayesian model averaging and deviance information criterion.  Modern computation via (Markov chain) Monte Carlo methods: Monte Carlo integration, sampling-importance resampling, Gibbs sampler, Metropolis-Hastings algorithms.  Mixture models, Hierarchical models, Bayesian regularization, Instrumental variables modeling, Large-scale (sparse) factor modeling, Bayesian additive regression trees (BART) and related topics, Dynamic models, Sequential Monte Carlo algorithms, Bayesian methods in microeconometrics, macroeconometrics, marketing and finance.

• Part I Bayesian ingredients: i) Inference: likelihood, prior, predictive and posterior distributions; ii) Model criticism: Marginal likelihoods, Bayes factor, model averaging and decision theory; and iii) Computation: An introduction (Markov chain and sequencial) Monte Carlo methods.
• Part II Multivariate models: i) Large-scale vector autoregressive models; ii) Factor models and other dimension reduction models; and iii) Time-varying high-dimensional covariance models.
• Part III Modern Bayesian statistical learning: i) Mixture models and the Dirichlet process: handling non-Gaussian models; ii) Regularization: sparsity via shrinkage and variable selection; iii) Large vector-autoregressive and factor models: combining sparsity and parsimony; iv) Classification and support vector machines; v) Regression trees and random forests; and vi) Latent Dirichlet allocation: Text as data, text mining.

Take-home midterm exam: 9am February 18th to 12pm February 20th. (data) (solution)

Homework assignments

1. HW1 – Due date: February 4th (12pm) (solution)
2. HW2 – Due date: February 11th (12pm) (solution)
3. HW3 – Due date: February 18th (9am) (solution + full conditionals + R code)
4. HW4 – Due date: March 25th (9am) – GPA data + some code: Use the GPA data to fit classical and Bayesian model choice strategies, similar to the ones we discussed in class for the Stock and Waton’s macro data, the Stine & Foster’s automation data and Wooldridge’s wage data.

Paper presentations:  Your task is to read the manuscript very carefully (several times, possibly) and prepare a 20-minute video presentation (no more than 25 minutos!) plus a concise 5-page summary of the manuscript (up to additional 5 pages for tables and graphs) to be sent to me no later than 12pm, April 8th 2021.

• Pamela: Çakmakl (2020) Modeling the density of US yield curve using Bayesian semiparametric dynamic Nelson-Siegel model, Econometric Reviews, 39(1), 71-91.
• Vanessa:  Gentzkow,  Kelly and Taddy (2019) Text as Data, Journal of Economic Literature, 57(3), 535-574.
• Eduardo: Pruser (2019) Forecasting with many predictors using Bayesian additive regression trees, Journal of Forecasting,  38(7), 621-631.
• Murilo:  Koop and Korobilis (2018) Variational Bayes inference in high-dimensional time-varying parameter models.
• Guilherme: Sezer, Gudelek and Ozbayoglu (2020) Financial time series forecasting with deep learning : A systematic literature review: 2005–2019.  Applied Soft Computing, 106181.

Examples developed in class: 

• Class of 01/14/2021: Slides from a more compact and more recent tutorial
• Class of 01/14/2021: Divergent priors, convergent posteriors
• Class of 01/14/2021: Bernoulli iid, logit regression or probit regression? (Rmarkdown)
• Class of 01/14/2021: Computing coronavac efficacy with logit, probit and clog-log generalized linear models?
• Class of 01/21/2021: Space Shuttle Challenger O-Ring – binary/glm binomial
• Class of 01/28/2021: R for mini-class 2: Bayesian AR(1) with Normal and t priors
• Class of 01/28/2021: R for Bayesian AR(2) with Normal and t priors
• Class of 01/28/2021: R for Bayesian AR(2) with Normal and t models
• Class of 01/28/2021: Bayesian multinomial model – MC integration & SIR
• Class of 01/28/2021: Bayesian binomial vs Poisson models – SIR (Analytical derivations – my handwriting)
• Class of 02/04/2021: Gibbs sampler – normal simple linear regression
• Class of 02/04/2021: Gibbs sampler – normal multiple linear regression
• Class of 02/04/2021: Bayesian AR(p) – conjugate analysis vs Gibbs sampler
• Class of 02/04/2021: Comparing MCMC strategies – Gibbs, MH, block/single
• Class of 02/04/2021: Limited dependent variable (Linear Tobit + data + R code)
• Class of 02/04/2021: Mixture of distributions (velocities of galaxies + data)
• Class of 02/11/2021: Nonlinear regression – comparing SIR and Gibbs+RWMH
• Class 0f 02/11/2021: Bimodal posterior: comparing random-walk MH and independent MH + R code
• Class 0f 02/11/2021:  Zero-Inflated Generalized Poisson model
• Class 0f 02/11/2021:  Markov chains and stationary/equilibrium densities
• Class 0f 02/11/2021:  Hamiltonian Monte Carlo: A toy example
• Class 0f 03/04/2021:  Normal linear models: subset selection via BIC and shrinkage/sparsity via regularizing priors – Stock and Watson’s (2002) macro data
• Class 0f 03/04/2021:  Normal linear models: k-fold cross-validation – Stine and Foster’s (2014) case study dataset – Rcode
• Class of 03/11/2021: More on linear models: In-sample vs out-of-sample MSE/MAE (R code +data)
• Class of 03/11/2021: Logistic model – ICU data: glm, bayesglm & glmnet (ridge, elastic net, lasso) (R code)
• Class of 03/11/2021: Receiver operating characteristics (ROC) and area under the curve (AUC)
• Class of 03/18/2021: Revisiting the ICU data: CART, BART and random forest (R code)
• Class of 03/25/2021: Revisiting Stock and Watson’s (2002) macro data

COURSE NOTES

PART I: Bayesian ingredients

1. Basic Bayes
2. Exchangeability
3. Principles of data reduction
4. More on estimators
5. Decision theory
6. Bayesian model criticism (pages 1-6 & 32-34)
7. Additional reading material:
   • Chapter 2 of Gamerman and Lopes (2006) – Compact, but easy to read.
   • Chapters 2-4 of Migon, Gamerman and Louzada (2014) – Integrates classical and Bayesian inference.
   • Chapter 1 and 2 of Gelman et al. (2013) – Application-oriented.
   • Chapter 4 (Sections 4.1-4.4) of Berger (1985) – More technical.
8. Discussion about p-values

PART II: Bayesian Computation

1. Monte Carlo (MC) methods
2. Markov chain: a brief review
3. Markov chain Monte Carlo (MCMC) algorithms
4. Hamiltonian Monte Carlo: A toy example
5. Stan/rstan for posterior inference: Hamiltonian MC (HMC) methods
6. MC and MCMC: Key References
7. More on Bayesian model criticism

PART III: Bayesian Learning

1. Fundamentos de Aprendizagem Estatística + R code + MC exercise
2. Multiple linear regression: selection, shrinkage, sparsity
   • Motorcycle example
   • Slides from the 2015 School of Time Series and Econometrics tutorial
   • Hahn, He and Lopes (2018) Gaussian linear regression with arbitrary sparsity + slides of a talk + R package bayeslm
   • Fava and Lopes (2021) The illusion of the illusion of sparsity + slides of a talk + UFPE webinar (62′ and forward)
   • Michael Betancourt’s Bayes Sparse Regression (stan/rstan example)
3. Classification: logistic regression and discriminant analysis
   • Sparse logistic regression for the spam/ham dataset (data)
   • Marketing campaigns of a Portuguese banking institution (data)
   • Sparse logistic regression: comparison of regularization and Bayesian implementations
   • Gelman, Jakulin, Pittau and Zu (2008) A weakly informative default prior distribution for logistic and other regression models, AOAS, 2(4), 1360-1383.
   • Polson, Scott and Windle (2013) Bayesian Inference for logistic models using Pólya-Gamma latent variables, JASA, 108, 1339-1349.
4. Bayesian factor analysis (BFA)
   • Bayesian factor analysis in R: a simple illustration (additional routines)
   • Factor models: an annotated bibliography (2003)
   • Lopes (2014)
5. Principal components analysis (PCA), PCA-based and FA-based regressions
6. Classification and regression trees (CART)
7. Bayesian CART
8. Bootstrap aggregating (bagging)
   • Bootstrap illustration
   • CART step-by-step for the Boston housing data
   • CART for the motorcycle data
9. Bayesian additive regression trees (BART)
10. Latent Dirichlet Allocation (LDA)
    • Twitter + BBC + Dickens, Wells, Verne or Austen? + The cat in the hat by Dr. Seuss
11. Neural Networks

Complementary material to PART III

1. Boosting (weak/stronger learners)
2. Random forests
3. Bayesian instrumental variables
4. General linear and hierarchical models
5. Limited dependent variable models
6. Finite mixture of distributions
7. Spatial models
8. P.Richard Hahn’s top 25 books on Statistics, Causal Inference, Statistical Computing, Machine Learning and Data Science

MATERIAL FROM PREVIOUS YEARS (2018-2020)

Homework assignments and take-home exams

1. Take-home midterm exam – 2020 (Bruno Levy’s solution + Igor Martins’ solution)
2. Take-home midterm exam – 2019
3. Take-home midterm exam – 2018 (Rafael Pucci’s solution + Raphael Gondo’s solution)
4. HW1 2019: Problems 2.26(a) and 2.26(c) (page 79), 3.1 (pages 106-107), 3.12 (page 110), 5.7 (page 185) and Example 5.1 (pages 143-146) – Gamerman and Lopes (2006) MCMC: Stochastic Simulation for Bayesian Inference. Errata of the numerator of the Bayes Factor of problem 2.26(a) + Errata of the denominator of the Bayes Factor of problem 2.26(a)
5. HW2 2019: Apply “Machine Learning” tools to the communities data from UCI Machine Learning Repository (http://archive.ics.uci.edu/ml/datasets.html).  See my simple exploratory analysis of the data as a start-up here.

Examples developed in class

1. Week of 01/13/2020: More on Bernoulli trials
2. Week of 01/13/2020: Monte Carlo integration and MC via importance sampling
3. Week of 01/20/2020: Student’s t: learning degree of freedom (Rmarkdown code)
4. Week of 01/20/2020: Student’s t regression (Rmarkdown code)
5. Week of 01/20/2020: Tobit linear regression (Dataset)
6. Week of 01/27/2020: Two-component mixture of univariate Gaussians
7. Week of 01/27/2020: SIR or Gibbs? The simple N-IG iid case
8. Week of 01/27/2020: Our first Metropolis-Hastings algorithm
9. Week of 01/27/2020: Linear model with the normal-gamma (NG) prior
10. Week of 01/27/2020: Linear model with NG prior: comparing MCMC schemes
11. Week of 02/03/2020: Bayesian skewed normal regression
12. Week of 02/03/2020: Model comparison via prior predictives: Normal vs t exercise
13. Week of 02/17/2020: A few R packages for Bayesian inference in linear models
14. Week of 02/17/2020: Discussion about the fallacies of p-values
15. Class of 01/15/2019: Posterior inference for the proportion of Bernoulli trial  (R code)
16. Week of 01/13/2019: Posterior of proportion of Bernoulli trial  (R code)
17. Class of 01/17/2019: Model comparison: Gaussian vs Student’s t
18. Class of 01/24/2019: Multivariate Gaussian: joint vs univariate Gibbs sampling
19. Class of 01/31/2019: Our 1st Metropolis-Hastings (MH) algorithm
20. Class of 01/31/2019: Our 2nd MH algorithm – comparing proposals
21. Class of 01/31/2019: Linear model with AR(1) errors (graphs)
22. Class of 03/19/2019: Human development index (by municipaliity in Brazil)
23. Class of May 3rd, 2018: Posterior inference for the variance in the normal case (R markdown code)
24. Class of May 8th, 2018: Bayesian model comparison (R markdown code)
25. Class of May 10th, 2018: Poisson model (Count data)(PDF with results)
26. Class of May 15th, 2018: Beta regression (PDF with results) + Ex 3.6-3.7 from GL(2006)
27. Class of May 17th, 2018: Comparing MC integration, SIR and Raoblackwellization 
28. Class of May 22nd, 2018: Bayesian linear regression
29. Class of May 24th, 2018: Sparse Bayesian linear regression (ridge, Bayesian lasso & horseshoe) – R code
30. Class of June 14th, 2018: My book chapter on Modern Bayesian Factor Analysis
31. Class of June 15th, 2018: Mixture of Poisson distributions (Rmd code)
32. Class of July 2nd, 2018: PCA and FA for term-structure data (dataset)
33. Class of July 10th, 2018: Bootstrap: sampling distribution of correlation coefficient

Bibliography: Bayesian econometrics

1. Zellner (1971) An Introduction to Bayesian Inference in Econometrics
2. Goel and Iyngar (1992) Bayesian Analysis in Statistics and Econometrics
3. West and Harrison (1997) Bayesian Forecasting and Dynamic Models (2nd edition)
4. Dorfman (1997) Bayesian Economics Through Numerical Methods
5. Bauwens, Lubrano and Richard (2000) Bayesian Inference in Dynamic Econometric Models
6. Koop (2003) Bayesian Econometrics
7. Geweke (2005) Contemporary Bayesian Econometrics and Statistics
8. Lancaster (2004) Introduction to Modern Bayesian Econometrics
9. Rossi, Allenby and McCulloch (2005) Bayesian Statistics and Marketing
10. Prado and West (2010) Time Series: Modeling, Computation and Inference
11. Geweke, Koop and Van Dijk (2011) The Oxford Handbook of Bayesian Econometrics
12. Greenberg (2013) Introduction to Bayesian Econometrics
13. Herbst and Schorfheide (2015) Bayesian Estimation of DSGE Models
14. Chan, Koop, Poirier and Tobias (2019) Bayesian Econometric Methods (2nd edition)
15. Broemeling (2019) Bayesian Analysis of Time Series
16. Bernardi, Grassi and Ravazzolo (2020) Bayesian Econometrics

Bibliography: Bayesian statistics

1. Berger (1985) Statistical Decision Theory and Bayesian Analysis
2. Bernardo and Smith (2000) Bayesian Theory
3. Gelman and Hill (2006) Data Analysis Using Regression and Multilevel/Hierarchical Models
4. Robert (2007) The Bayesian Choice: From Decision-Theoretic Foundations to Computational Implementation
5. Hoff (2009) A First Course in Bayesian Statistical Methods
6. Carlin and Louis (2009)  Bayesian Methods for Data Analysis (3rd edition)
7. Gelman, Carlin, Stern, Dunson, Vehtari and Rubin (2016) Bayesian Data Analysis
8. Migon, Gamerman and Louzada (2015) Statistical Inference: An Integrated Approach (2nd edition)
9. Reich and Ghosh (2019) Bayesian Statistical Methods
10. Held and Sabanes-Bove (2020) Likelihood and Bayesian Inference: With Applications in Biology and Medicine

Bibliography: Bayesian computation

1. Gilks, Richardson and Spiegelhalter (1995) Markov Chain Monte Carlo in Practice
2. Doucet, de Freitas and Gordon (2001) Sequential Monte Carlo Methods in Practice
3. Robert and Casella (2004) Monte Carlo Statistical Methods (2nd edition)
4. Gamerman and Lopes (2006) MCMC: Stochastic Simulation for Bayesian Inference, Second Edition
5. Marin and Robert (2007) Bayesian Core: A Practical Approach to Computational Bayesian Statistics
6. Albert (2009) Bayesian Computation with R
7. Brooks, Gelman, Jones and Meng (2011) Handbook of Markov Chain Monte Carlo
8. Givens and Hoeting (2012) Computational Statistics (2nd edition)
9. Marin and Robert (2014) Bayesian Essentials with R (complete solution manual)
10. Turkman, Paulino and Mueller (2019) Computational Bayesian Statistics: An Introduction
11. McElreath (2020) Statistical Rethinking: A Bayesian course with Examples in R and STAN
12. Chopin and Papaspiliopoulos (2020) An Introduction to Sequential Monte Carlo

Bibliography: (Bayesian) statistical learning

1. Bishop (2006) Pattern Recognition and Machine Learning
2. Hastie, Tibshirani and Friedman (2008) The Elements of Statistical Learning, 2nd edition
3. Murphy (2012) Machine Learning: A Probabilistic Perspective
4. Barber (2012) Bayesian Reasoning and Machine Learning
5. James, Witten, Hastie and Tibshirani (2013) An Introduction to Statistical Learning
6. Hastie, Tibshirani and Wainwright (2015) Statistical Learning with Sparsity
7. Efron and Hastie (2016) Computer Age Statistical Inference: Algorithms, Evidence and Data Science
8. Fernandez and Marques (2018) Data Science, Marketing and Business
9. Izbicki & Santos (2020) Aprendizado de máquina: uma abordagem estatística

Bibliography: Classical Monte Carlo papers

• Metropolis and Ulam (1949) The Monte Carlo method. JASA, 44, 335-341.
• Metropolis, Rosenbluth, Rosenbluth, Teller and Teller (1953) Equation of state calculations by fast computing machines. Journal of Chemical Physics, 21, 2087-1092.
• Hastings (1970) Monte Carlo sampling methods using Markov chains and their applications. Biometrika, 57, 97-109.
• Peskun (1973) Optimum Monte Carlo sampling using Markov chains. Biometrika, 60, 607-612.
• Besag (1974) Spatial Interaction and the Statistical Analysis of Lattice Systems. JRSS-B, 36, 192-236.
• Kirkpatrick, Gelatt and Vecchi (1983) Optimization by Simulated Annealing. Science, 220 (4598), 671-680.
• Geman and Geman (1984) Stochastic relaxation, Gibbs distributions, and the Bayesian restoration of images. IEEE Trans. Pattern Analysis and Machine Intelligence, 6, 721-741.
• Pearl (1987) Evidential reasoning using stochastic simulation of causal models. Artificial intelligence, 32, 245-257.
• Tanner and Wong (1987) The Calculation of Posterior Distributions by Data Augmentation. JASA, 82, 528-540.
• Geweke (1989) Bayesian Inference in Econometric Models Using Monte Carlo Integration. Econometrica, 57, 1317-1339.
• Gelfand and Smith (1990) Sampling-Based Approaches to Calculating Marginal Densities. JASA, 85, 398-409.
• Casella and George (1992) Explaining the Gibbs Sampler. The American Statistician, 46, 167-174.
• Gilks and Wild (1992) Adaptive Rejection Sampling for Gibbs Sampling. Applied Statistics, 41, 337-348.
• Smith and Gelfand (1992) Bayesian Statistics without Tears: A Sampling-Resampling Perspective. The American Statistician, 46, 84-88.
• Chib and Greenberg (1995) Understanding the Metropolis-Hastings algorithm. The American Statistician, 49, 327-335.

Course: ECONOMETRICS III 2021 (TIME SERIES) – Doctoral Program in Business Economics
Professor: Hedibert Freitas Lopes – www.hedibert.org

Objective: The main goal of the course is to make the student familiar with and able to implement univariate and multivariate time series models by using both frequentist and Bayesian approaches. All classroom examples and implementations as well as projects will be carried out by the open-source statistical software R.

Course description: Brief review of frequentist inference followed by the introduction of key ingredients of Bayesian inference, model selection and criticism. An introduction to the main Monte Carlo methods for Bayesian inference: MC integration, resampling, MCMC and sequential MC. Univariate time series models, including AR(F)IMA models, state-space models, Markov switching models, GARCH and stochastic volatility models. Multivariate time series models, including Bayesian VARs and factor-augmented VARs, dynamic factor models, time-varying covariance models.

• PART I: Basic univariate time series models: AR, MA and ARMA models; Unit-root non-stationarity and long-memory processes; Seasonal models.
• PART II: Bayesian ingredients (prior, likelihood, posterior, predictive, Bayes factor and posterior model probability); Monte Carlo (MC)  methods (MC integration, sampling importance resampling (SIR)) and Markov chain Monte Carlo (MCMC) methods (Gibbs sampler and Metropolis-Hastings (MH) algorithms).
• PART III: More univariate time series: ARCH/GARCH models; EGARCH, GARCH-M, TGARCH; Bayesian GARCH; Bayesian inference in the local level model; Dynamic models; Stochastic volatility models.  We will use MCMC as well as sequential Monte Carlo (SMC) schemes to perform batch and online posterior inference.
• PART IV: Multivariate time series models: Vector autoregressive (VAR) models; Large Bayesian VAR (BVAR) models, factor augmented VAR (FAVAR) models, time-varying parameter BVAR (TVP-BVAR) models, Bayesian FAVAR (BFAVAR) models; Factor models and time-varying covariance models.

Bibliography

• Gamerman and Lopes (2006) MCMC: Stochastic Simulation for Bayesian Inference, Second Edition. Chapman & Hall/CRC. http://www.dme.ufrj.br/mcmc
• Tsay (2010) Analysis of Financial Time Series, Third Edition. Wiley-Interscience, Probability and Statistics. http://faculty.chicagobooth.edu/ruey.tsay/teaching/fts3
• Tsay (2014) Multivariate Time Series Analysis with R and Financial Applications. Wiley. http://faculty.chicagobooth.edu/ruey.tsay/teaching/mtsbk
• Shumway and Stoffer (2011) Time Series Analysis and Its Applications with R Examples, Third Edition. Springer.  http://www.stat.pitt.edu/stoffer/tsa4

Teaching assistant: Bruno do Prado Costa Levy (brunopcl at al dot insper dot edu dot br) – Wednesdays from 10am to 11:30am

Take-home midterm exam: 9am of Tuesday, February 23th to 12pm of Thursday, February 25th. (data)

Homework assignments

1. HW1 – Due date: January 26th – Problems 1,2,3,8,19,20 and 21, chapter 1 of Shumway and Stoffer’s book (4th edition).
   • HW1 by Alexandre + HW1 by Livia + HW1 by Victoria
2. HW2 – Due date: February 2nd – Problem 2.15 (page 107) of Tsay’s (2010) book.  However, download the up-to-date quarterly gross domestic implicit price deflator time series from the Federal Reserve Bank of St Louis.  Fit the ARIMA models with data up to the 4th quarter of 2018 and use 2019.I to 2020.III (7 quarters) for forecasting comparisons.
   • HW2 by Thaline
3. HW3 – Due date: Tuesday, February 23rd 2021 at 9am
4. HW4 – Due date: Tuesday, March 9th 2021 at 9am – Fit Gaussian and Student’s t GARCH(1,1) to Vale S.A. (VALE) using the R packages garchFit, bayesGARCH and RSTAN that I have provided when we studied Petrobras (PBR).  Feel free to add other (non-Bayesian) GARCH-type fits based on the ARCH-glossary that we have discussed in class.
5. HW5 – Due Date: Tuesday, March 23rd 2021 at 9am – Collect meaningful time-series (suggestion: Apple & SP500 for US market or Vale & Ibovespa for the Brazilian market).  Use the time series in a simple CAP-M model which allows for  time-varying slope.  Repeat it by allowing ONLY time-varying slope.  Then, allow both time-varying intercept and slope.  Compare the three models to the benchmark OLS fit.  You can use whatever Bayesian R package (such as bsts) that produces posterior summaries of the latent variables (intercepts and slopes) as well as static parameters.  Comment your findings.  There are lots of papers out there, and I recommend a 16-year old one here: Jostova and Philipov (2005) Bayesian analysis of stochastic betas, Journal of Financial and Quantitative Analysis, 747-778.

Paper presentations: Prepare a 20min video plus 5-page summary of the paper to be sent to me and Bruno by 12pm on Tuesday, April 13th 2021.

1. Pedro: Graziadei, Lopes and Marques (2020) Bayesian generalizations of the integer-valued autoregressive model, Journal of Applied Statistics. 
2. Victoria: Silva, Lopes and Migon (2006) The extended generalized inverse Gaussian distribution for log-linear and stochastic volatility models, Brazilian Journal of Probability and Statistics, 67-91.
3. Alexandre: Carvalho and Lopes (2006) Simulation-based sequential analysis of Markov switching stochastic volatility models, Computational Statistics and Data Analysis, 51 (9), 4526-4542.
4. Vinicius: Warty, Lopes and Polson (2018) Sequential Bayesian learning for stochastic volatility with variance-gamma jumps in returns, Applied Stochastic Models in Business and Industry, 2018, 34, 460-483.
5. Nathalia: Prado and Lopes (2013)  Sequential parameter learning and filtering in structured autoregressive state-space models, Statistics and Computing, 23 (1), 43-57.
6. Livia: Primiceri (2005) Time Varying Structural Vector Autoregressions and Monetary Policy, The Review of Economic Studies, Vol. 72, No. 3, 821-852.
7. Thaline: Carriero, Todd and Massimiliano (2019)  Large Bayesian vector autoregressions with stochastic volatility and non-conjugate priors, Journal of Econometrics, 212(1), 137-154.
8. Guilherme: Shirota, Omori, Piao and Lopes (2017) Cholesky realized stochastic volatility model, Econometrics and Statistics, 2017, 3, 34-59.
9. Giovanna: Kastner, Fruehwirth-Schnatter and Lopes (2017) Efficient Bayesian inference for multivariate factor SV models, Journal of Computational and Graphical Statistics, 26, 905-917.
10. Rafael: Kastner and Huber (2020) Sparse Bayesian vector auto-regressions in huge dimensions, Journal of Forecasting, 30(7), 1142-1165.
11. Thayla: Lopes, McCulloch and Tsay (2020) Parsimony inducing priors for large scale state-space models, Journal of Econometrics (Revised & Resubmitted).
12. Octavio: Levy and Lopes (2021) Dynamic ordering learning in multivariate forecasting.

Examples developed in class

1. Class of 01/12/2021 – Brief introduction to time series in R
2. Class of 01/19/2021 – AR(1), random walk and AR(p) models
3. Class of 01/19/2021 – ARMA & ARIMA models
4. Class of 01/26/2021 – ARFIMA models
5. Class of 01/26/2021 – Bayesian AR(1)
6. Class of 02/02/2021 – Bayesian AR(1) with Normal and t priors
7. Class of 02/02/2021 – Bayesian AR(2) with Normal and t priors
8. Class of 02/02/2021 – Bayesian AR(2) with Normal and t models
9. Class of 02/02/2021 – Bayesian nonlinear regression – SIR and RW-MH
10. Class of 02/09/2021 – Bayesian AR(p) – conjugate analysis vs Gibbs sampler
11. Class of 02/09/2021 – Comparing MCMC strategies – Gibbs, MH, block/single
12. Class of 02/09/2021 – Nonlinear regression – comparing SIR and Gibbs+RWMH
13. Class 0f 02/09/2021 – Bimodal posterior: comparing random-walk MH and independent MH + R code
14. Class of 02/09/2021 – Linear Gaussian regression with Normal-Half-Cauchy prior – MCMC with Gibbs and  RWMH  steps + R code
15. Class of 03/02/2021 – Petrobras (PBR): ARCH(1,1) + GARCH(1,1)
16. Class of 03/02/2021 – Petrobras (PBR): garchFit – bayesGARCH – rstan (stan file)
17. Class of 03/02/2021 – Modeling S&P 500 realized volatility & log returns (stan code + graphs + data)
18. Class of 03/09/2021 – Modeling COVID-19 death: an exercise in state-space modeling
19. Class of 03/09/2021 – Hamilton’s (2017) paper “Why you should never use the HP filter”
20. Class of 03/23/2021 – SV-AR(1) for PBR: MCMC, SMC/particle filter and sequential MCMC (data)
21. Class of 04/06/2021 – Univariate stochastic volatility, factor analysis, & factor stochastic volatility
22. Class of 04/06/2021 – Bayesian time-varying covariance: DCC and FSV models (R code)

TEACHING MATERIAL

PART I: Basic univariate time series

1. Autoregressive (AR) models and moving average (MA) models (HTML output)
2. Unit-root nonstationarity and long-memory processes  (HTML output)
3. Seasonal models
   • SARIMA in R – Brazilian industrial production
   • R code for airline data – ARIMA(0,1,1)(0,1,1)[12] (R markdown html output)

PART II: Basic Bayes

1. Bayesian ingredients
2. Monte Carlo (MC) methods
3. Markov chain Monte Carlo (MCMC) methods
4. Using stan/rstan for approximate Bayesian inference via Hamiltonian MC (HMC) methods
5. MC and MCMC: Key References

PART III: Garch-type, dynamic linear and stochastic volatility models  – MCMC and SMC

1. ARCH/GARCH-type models
   • Glossary of ARCH models
   • EGARCH, GARCH-M, TGARCH
   • Bayesian GARCH
2. Dynamic linear models (DLMs)
   • GARCH(1,1) versus stochastic volatility AR(1): motivating dynamic modeling (R code)
   • Local level model: O(n^2) updates versus O(n) updates (the FFBS) (R code)
   • Normal dynamic linear regression: Kalman, FFBS/MCMC and sequential MCMC (Rcode)
   • Normal dynamic linear regression: block sampling (FFBS) vs component-wise sampling (Rcode)
   • Toy DLM + Another toy DLM + AR1 + noise DLM
3. Nonlinear dynamic model – MCMC sampling individual states conditional on all other states
4. Sequential Monte Carlo – pure filter
   • Bootstrap filter (BF)
   • BF and sequential importance sampling (SIS)
   • BF, auxiliary particle filter (APF), optimal particle filter (OPF)
5. Sequential Monte Carlo – parameter learning
   • SMC with parameter learning
6. Stochastic volatility models
   • MCMC for the SV-AR(1) model
   • BF for the SV-AR(1) model
7. Using R packages “stochvol” & “rstan” for SV with Gaussian or Student’s t errors

PART IV: Multivariate time series

1. Vector autoregressive models
   • R code for VAR, Chapter 2 of Tsay MTS book + packages
2. Large BVAR, FAVAR, TVP-BVAR & BFAVAR
   • Koop ad Korobilis (2009) Bayesian MTS Methods for Empirical Macroeconomics.
3. Factor models (Standard factor analysis, Spatial dynamic factors, Factor stochastic volatility)
4. Time-varying covariance models
   • Efficient Bayesian inference for multivariate FSV models (Kastner, Fruewirth-Schnatter & Lopes)
   • Factor stochastic volatility with time varying loadings and Markov switching regimes (Lopes & Carvalho)
   • Cholesky realized SV models  (Shirota, Omori, Lopes & Piao)
   • Bayesian inference for stochastic volatility modeling (Lopes & Polson)
   • A Review of Stochastic Volatility: univariate and multivariate models (Platanioti, McCoy & Stephens)

Additional reading material on Bayesian time series

1. Bayesian Statistics (a very brief introduction) – Ken Rice, April, 2014
2. Lopes and Salazar (2006) Bayesian model uncertainty in smooth transition autoregressions, Journal of Time Series Analysis, 27, 99-117.
3. Huerta and Lopes (2000) Bayesian forecasting and inference in latent structure for the Brazilian industrial production index, Brazilian Review of Econometrics, 20, 1-26.
4. Kleibergen and Hoek (2000) Bayesian Analysis of ARMA Models. Tinbergen Institute Discussion Paper.
5. Marriott, Ravishanker, Gelfand and Pai (1995) Bayesian Analysis of ARMA Processes: Complete Sampling Based Inference under Exact Likelihoods. Bayesian Statistics and Econometrics: Essays in honor of Arnold Zellner. Berry, Chaloner and Geweke, eds., John Wiley & sons, 241-254.

R stuff

Radford Neal’s 13 lectures about R

McLeod, Yu and Mahdi’s (2012) Time Series Analysis with R

MATERIAL FROM PREVIOUS YEARS (2018-2020)

Homework assignments and take-home exams

1. Take-home midterm exam 2020
2. Take-home midterm exam 2019
3. Midterm exam 2017: solution
4. HW1 2020
5. HW2 2020
6. HW3 2020: R code
7. HW4 2020: VAR e BVAR para dados de consumo de energia em 7 estados brasileiros (R code)
8. HW1 2019: Simple MA model + Exercises 1.1 to 1.5 and 2.1 to 2.4 of Tsay (2010).
9. HW2 2019: dataset
10. HW3 2019: VAR and BVAR (problems 2.4, 2.5 and 2.6 of Tsay (2014) Multivariate Time Series)
11. HW2 2018: Turn in exercises 3.1 and 3.2 of Hamilton (1994)  and exercises 2.7, 2.8 and 2.9 of Tsay (2010).
12. HW3 2018: MC Integration, SIR & Gibbs sampler
13. HW1 2017: Solution to  2b and 2c + Additional MC exercise + Solution to 2d.
14. HW2 2017: dataset

Paper presentations

1. Del Negro and Schorfheide (2004) Priors from General Equilibrium Models for VARS. IER, 45, 643-673.
2. Banbura, Giannone and Reichlin (2010) Large BVARs, JAE, 25(1), 71-92.
3. Koop and Korobilis (2013) Large TVP VARs. JoE, 177, 185-198.
4. Giannone, Lenza and Primiceri (2015) Prior selection for VARs. The Review of Economics and Statistics, 97,436-451.
5. Carriero, Clark and Marcellino (2015) BVARs: Specification choices and forecast accuracy. JAE, 30, 46-73.
6. Chan and Eisenstat (2018) Bayesian model comparison for TVP VARs with SV. JAE, 33, 509-532.
7. Carriero, Clark and Marcellino (2019) Large BVARs with SV and flexible priors. JoE, 212, 137-154.
8. Kastner and Huber (2020) Sparse Bayesian vector auto-regressions in huge dimensions, JoF, 30(7), 1142-1165.
9. Korobilis and Pettenuzzo (2020) Adaptive hierarchical priors for high-dimensional VARs. JoE, 212(1), 241-271.
10. Koop, Korobilis and Pettenuzzo (2019) Bayesian Compressed VARs. JoE, 210, 135-154.

Examples developed in class

1. Week of 01/13/2020:  ACF of white noise and random walk processes
2. Week of 01/13/2020: AR(3) simulation exercise(R markdown code)
3. Week of 01/20/2020: AR(1) models: predictive analysis  (R markdown code)
4. Week of 01/27/2020: AR(3) models: Gibbs sampler (html) (Rmarkdown code )
5. Week of 01/27/2020: Our first Metropolis-Hastings algorithm
6. Week of 01/27/2020: Bayesian regression with the normal-gamma prior
7. Class of 01/15/2019: ACF of white noise and random walk processes
8. Class of 01/22/2019: AR(3) simulation exercise(R markdown code)
9. Class of 01/24/2019: Gaussian and non-Gaussian GARCH models + Rmarkdown  + Petrobras data
10. Class of 01/29/2019: Our first state-space model: AR(1) plus noise model
11. Class of 01/31/2019: Linear regression with AR(1) errors (graphs)
12. Class of 02/05/2019: AR(1) plus noise model: FFBS
13. Class of 02/07/2019: AR(1) plus noise model: block-move vs single-move
14. Class of 02/26/2019: My first particle filter
15. Class of 02/28/2019: SV-AR(1): MCMC & SMC  + (R code)
16. Class of 03/01/2019: SV & FSV  + (R code) + (dados)
17. Class of 03/26/2019: DCC-GARCH & FSV  + (R code)
18. Class of April 19th, 2018: ACF of white noise and random walk processes
19. Class of May 3rd, 2018: AR(3) simulation exercise(R markdown code)
20. Class of May 8th, 2018: Sampling distribution of the Dickey-Fuller ratio
21. Class of May 11th, 2018: SARIMA for unemployment rate in Sao Paulo (R code)
22. Class of May 15th, 2018: Sequential Bayesian learning
23. Class of May 17th, 2018: Monte Carlo integration/simulation
24. Class of May 22nd, 2018: Gaussian AR(2) model with conditionally conjugate priors: Gibbs Sampler
25. Class of May 24th, 2018: Bayesian linear regression
26. Class of May 24th, 2018: Binomial model and mixture of betas prior: comparing SIR and Metropolis-Hastings schemes
27. Class of May 25th, 2018: Bayesian CAPM
28. Class of May 29th, 2018: Gaussian and non-Gaussian GARCH models + Rmarkdown  + Petrobras
29. Class of June 5th, 2018:  Modeling time-varying variances via stochastic volatility (SV) models
30. Class of June 14th, 2018: Hidden Markov model: forward filtering, backward sampling (Rmd code)
31. Class of June 21st, 2018: VAR homework (Due date: July 5th 10:30 am)
32. Class of April 18th, 2017: R code for the AR(2) example worked in class
33. Class of April  18th, 2017: R for Shumway and Stoffer’s chapter 1
34. Class of April  18th, 2017: More R code for the AR(1) and AR(2) processes  (Slides)
35. Class of April  18th, 2017: R markdown script (run via Rstudio) (PDF output or HTML output)
36. Class of April 25th, 2017: Monte Carlo exercise: studying the sampling behavior of the t test under unit root
37. Class of May 1st, 2017: Bayesian inference for the Gaussian AR(2) model (R code)
38. Class of May 8th, 2017: Computing pi via rejection sampling: our first MC sampling scheme
39. Class of May 8th, 2017: MC integration for a simple normal-normal example
40. Class of May 8th, 2017: Gibbs sampler for AR(1) model with a changing point (changing in the intercept)
41. Class of May 16th, 2017: Brazilian monthly production of cement (January 2002 to February 2017)
42. Class of May 16th, 2017: AR(1) plus noise + Figure 1 +Figure 2.
43. Class of May 16th, 2017: AR(1) plus noise – Kalman filter and smoother + Figures.
44. Class of May 16th, 2017: AR(1) plus noise – Bayesian inference via MCMC/FFBS + Figures.
45. Class of May 16th, 2017: AR(1) plus noise – Modeling Alcoa realized volatilities via 1st order DLM + Data.
46. Class of May 24th, 2017: AR(1) plus noise – Comparing block move (FFBS) with single move MCMC schemes
47. Class of June 6th, 2017: Linear regression with Markov switching intercept – R code + Figures

Course: TIME SERIES (Spring 2023)

Professor: Hedibert Freitas Lopes – www.hedibert.org

Lectures: Mondays and Wednesdays, from 10:30am to 11:45am (January 10th to April 27th)

Office hours: Wednesdays, from 11am to 12pm (by appointment only)

Classroom: Social Sciences 205

ATTENTION: This is an advanced time series course (see course description below!). A strong background in calculus, probability, statistics and matrix algebra is highly beneficial. 

Syllabus

Course description: The main goal of the course is to make the student familiar with and able to implement univariate and multivariate modern time series models. Univariate time series models we will consider include the family of autoregressive (fractionally) integrated moving average (ARIMA) models, dynamic linear models (aka state-space) models, Markov switching models, generalized autoregressive conditionally heteroskedastic (GARCH) and stochastic volatility (SV) models. Multivariate time series models we will considere include vector autoregressive (VAR) models, factor-augmented VARs, dynamic factor models and various time-varying covariance models. The inferential approach of this course is predominantly Bayesian, so we will briefly introduce key ingredients of Bayesian inference, model selection and criticism. An introduction to the main Monte Carlo methods for Bayesian inference, such as MC integration, sampling-importance-resampling (SIR), Markov chain Monte Carlo (MCMC) and sequential MC (SMC), will also be introduced. All classroom examples and implementations as well as projects will be carried out by the open-source statistical software R.

Key topics covered will be:

• PART I: Basic univariate time series models: AR, MA and ARMA models; unit-root non-stationarity and long-memory processes; seasonal models.
• PART II: Bayesian ingredients (prior, likelihood, posterior, predictive, Bayes factor and posterior model probability); Monte Carlo (MC) methods (MC integration, sampling importance resampling (SIR)) and Markov chain Monte Carlo (MCMC) methods (Gibbs sampler and Metropolis-Hastings (MH) algorithms).
• PART III: More univariate time series: ARCH/GARCH models; EGARCH, GARCH-M, TGARCH; Bayesian GARCH; Bayesian inference in the local level model; Dynamic models; Stochastic volatility models. We will use MCMC as well as sequential Monte Carlo (SMC) schemes to perform batch and online posterior inference.
• PART IV: Multivariate time series models: Vector autoregressive (VAR) models; Large Bayesian VAR (BVAR) models, factor augmented VAR (FAVAR) models, time-varying parameter BVAR (TVP-BVAR) models, Bayesian FAVAR (BFAVAR) models; Factor models and time-varying covariance models.

Useful textbooks:

• Gamerman and Lopes (2006) MCMC: Stochastic Simulation for Bayesian Inference, Second Edition. Chapman & Hall/CRC. http://www.dme.ufrj.br/mcmc
• Prado, Ferreira and West (2021) Time Series: Modeling, Computation & Inference, Second Edition. Chapman & Hall/CRC.
• Shumway and Stoffer (2011) Time Series Analysis and Its Applications with R Examples, Third Edition. Springer. http://www.stat.pitt.edu/stoffer/tsa4
• Tsay (2010) Analysis of Financial Time Series, Third Edition. Wiley-Interscience, Probability and Statistics. http://faculty.chicagobooth.edu/ruey.tsay/teaching/fts3
• Tsay (2014) Multivariate Time Series Analysis with R and Financial Applications. Wiley. http://faculty.chicagobooth.edu/ruey.tsay/teaching/mtsbk

Homework assignments:  HW are to be delivered via a single PDF file to hedibert@gmail.com

• HW1 – Due date: At the beginning of class on Tuesday, January 31st, 2023
• HW2 – Due date: At the beginning of class on Thursday, February 23rd, 2023.
• HW3 – Due date: At the beginning of class on Thursday, March 16th, 2023.
  • Fit Gaussian and Student’s t GARCH(1,1) to your favorite returns (Coke, Apple, Amazon, S&P, etc) using the R packages garchFit and bayesGARCH that I have used in class.  Feel free to add other (non-Bayesian) GARCH-type fits based on the ARCH-glossary that we have discussed in class.  Use data between January 2005 and December 2022, so you are including the 2007-2008 financial crisis, as well as the 2020-2021 COVID pandemic.  Comment your findings.

• HW4 – Due date: At the beginning of class on Thursday, April 6th, 2023.
  • Inspired by HW3 (above), fit Gaussian and Student’s t SV-AR(1) models, as well their extended versions that contemplate leverage effect (skewed effect between large positive returns and large negative returns), to your favorite returns (Coke, Apple, Amazon, S&P, etc) using the R packages stochvol (by Gregor Kastner).  Use data between January 2005 and December 2022, so you are including the 2007-2008 financial crisis, as well as the 2020-2021 COVID pandemic.  Comment your findings, including comparisons with the GARCH-type models from HW3.  Hint:  We basically perform this task in Section 5 of the following example: sv-ar(1) for S&P500 returns.  Have fun! 

List of papers for final presentation – Due date: 12pm, May 2nd 2023.

Your final evaluation has two parts AND both need to be turned on May 2nd 2023 at noon: a) A five-page summary of the paper, and b) recorded 15-minute presentation of the paper plus slides.

1. Stefano Chiaradonna – Cyber risk measurement via loss distribution approach and GARCH model, Communications for Statistical Applications and Methods, 2023, Vol. 30, No. 1, 75–94.  By Sanghee Kim and Seongjoo Song. https://doi.org/10.29220/CSAM.2023.30.1.075
2. John Schiele – On the long run volatility of stocks: time-varying predictive systems, Journal of the American Statistical Association, 2018, 113, 1050-1069. By Carlos Carvalho, Hedibert Lopes & Robert McCulloch.
3. Lydia Gabric – Bayesian prediction of risk measurements using copulas, in Bocker, K. (Ed.) Rethinking Risk Measurement and Reporting: Uncertainty, Bayesian Analysis and Expert Judgement, 2010, 553-578. By Ausin and Lopes.  https://hedibert.org/wp-content/uploads/2013/12/ausin-lopes-2010.pdf
4. Shuai Zhu – Bayesian generalizations of the integer-valued autoregressive model, Journal of Applied Statistics.  By Graziadei, Lopes and Marques (2020)
5. Chukwudi Obite – Simulation-based sequential analysis of Markov switching stochastic volatility models, Computational Statistics and Data Analysis, 51 (9), 4526-4542. By Carvalho and Lopes (2006) 
6. Fan Wu – Time Varying Structural Vector Autoregressions and Monetary Policy, The Review of Economic Studies, Vol. 72, No. 3, 821-852. By Primiceri (2005)
7. Xianjian Xie – Sparse Bayesian vector auto-regressions in huge dimensions, Journal of Forecasting, 30(7), 1142-1165. By Kastner and Huber (2020) 

TEACHING MATERIAL

PART I: Basic univariate time series

1. Autoregressive (AR) models and moving average (MA) models (HTML output)
   • R code: Brief introduction to time series in R
   • R code: AR(1), random walk and AR(p) models
   • R code: ARMA & ARIMA models
2. Unit-root nonstationarity and long-memory processes  (HTML output)
3. Seasonal models
   • SARIMA in R – Brazilian industrial production
   • R code for airline data – ARIMA(0,1,1)(0,1,1)[12] (R markdown html output)

PART II: Basic Bayes

1. Bayesian ingredients
   • Gaussian vs Cauchy model & Gaussian prior
   • Gaussian linear regression (Koop and Tobias dataset)
   • Bayesian AR(1) with conjugate prior
2. Bayesian computation
   • Monte Carlo (MC) methods
     • Bayesian AR(1)
     • Bayesian AR(1) with Normal and t priors
     • Bayesian AR(2) with Normal and t priors
     • Bayesian AR(2) with Normal and t models (HTML)
     • Bayesian regression with autocorrelated errors
     • SIR, scale mixture of normals and raoblackwellization
     • Sampling from the log-chi-square distribution – SIR
   • Markov chain Monte Carlo (MCMC) methods
     • Bivariate normal – Gibbs sampler
     • Bayesian AR(p) – conjugate analysis vs Gibbs sampler
     • Random walk Metropolis: two toy examples
     • Comparing MCMC strategies – Gibbs, MH, block/single
     • Revisiting regression with autocorrelated errors: SIR vs Gibbs
     • Threshold AR (TAR) model: Gibbs and Metropolis steps
   • MC and MCMC: Key References

PART III: Garch-type, dynamic linear and stochastic volatility models

1. Glossary of ARCH models
   • EGARCH, GARCH-M, TGARCH
   • Bayesian GARCH
2. Dynamic models
   • Dynamic linear regression – Rmarkdown
   • A few review papers:
     • Migon, Gamerman, Lopes and Ferreira (2005)
     • Schmidt and Lopes (2019)
     • Migon, Alves, Menezes and Pinheiro (2023)
   • Example 1: Local level model: fixed variances
   • Example 2: Simple dynamic regression: fixed variances
   • Example 3: Simple dynamic regression: learning variances
   • Example 4: BSTS package for state-space modeling of NO3 (Data)
   • Steve Scott’s BSTS tutorial (my own shorter tutorial)
   • Hidden Markov Model: Variance Switching
   • Stochastic volatility models
   • Petrobras example: SV, SVt, SVl & SVtl
3. Sequential Monte Carlo – pure filter
   • Local level DLM & nonlinear DM
   • Stochastic volatility: MCMC vs SMC
   • Stochastic volatility: Particle filter (pure filters) vs Brute force MCMC

PART IV: Multivariate time series

1. Vector autoregressive models (VAR) part one
   • VAR and BVAR in R – a couple of examples
   • Minnesota prior for the trivariate BVAR(2)
2. VAR part two: Large BVAR, FAVAR, TVP-BVAR & BFAVAR
3. Bayesian factor analysis (BFA)
   • Bayesian factor analysis in R: a simple illustration (additional routines)
   • Factor models: an annotated bibliography (2003)
   • Lopes (2014)
   • Factor modeling of pollutant NO3 (Data)
4. Time-varying covariance modeling
   • Factor stochastic volatility models
   • Efficient Bayesian inference for multivariate FSV models (Kastner, Fruewirth-Schnatter & Lopes)
   • Factor stochastic volatility with time varying loadings and Markov switching regimes (Lopes & Carvalho)
   • Cholesky realized SV models  (Shirota, Omori, Lopes & Piao)
   • Bayesian inference for stochastic volatility modeling (Lopes & Polson)
   • A Review of Stochastic Volatility: univariate and multivariate models (Platanioti, McCoy & Stephens)
   • An bivariate example

Bonus topic: Time series meet machine learning

• Deep Learning Models For Inflation Forecasting (Theoharidis, Guillen and Lopes)
• Forecasting Inflation in a Data-Rich Environment: The Benefits of Machine Learning Methods (Medeiros, Vasconcelos, Veiga and Zilberman)
• Real-time inflation forecasting with high-dimensional models (Garcia, Mederios and Vasconcelos)
• Adaptive models and heavy tails with an application to inflation forecasting (Monache and Petrella)
• Deep learning with long short-term memory networks for financial market predictions (Fischer and Krauss)
• A comparison of time series and machine learning models for inflation forecasting. Empirical evidence from the USA (Sahin and Subasi)
• Empirical Asset Pricing via Machine Learning (Gu, Kelly, Xiu)

Old homework assignments (spring 2022): HW1 (Solution) + HW2  + HW3 (Derivations + R code) + HW4: Fit Gaussian and Student’s t GARCH(1,1) to your favorite returns (Coke, Apple, Amazon, S&P, etc) using the R packages garchFit and bayesGARCH that I have used in class.  Feel free to add other (non-Bayesian) GARCH-type fits based on the ARCH-glossary that we have discussed in class.  Use data between January 2005 and December 2021, so you are including the 2007-2008 financial crisis, as well as the 2020-2021 COVID pandemic.  Comment your findings.

Course: ADVANCED BAYESIAN STATISTICAL LEARNING (Spring 2023)

Professor: Hedibert Freitas Lopes – www.hedibert.org

Lectures: Tuesdays and Thursdays, from 9:00am to 10:15am (January 10th to April 27th)

Office hours: Wednesdays, from 10am to 11am (by appointment only)

Classroom: Social Sciences 205

ATTENTION: This is an advanced Bayesian course (see course description below!). A strong background in calculus, probability, statistics and matrix algebra is highly beneficial. 

Syllabus

Course description: The end of the course goal is to expose the student to modern Bayesian solutions to highly structured and stochastic real world problems.  We will visit well known Bayesian issues, such as prior specification/sensitivity, model comparison/criticism and model averaging, as well as Bayesian computation via various Monte Carlo methods.  We approach regularization in linear and log-linear models via Bayesian LASSO, Spike-and-Slab priors and related sparsity-inducing priors.  We cover decoupling shrinkage and selection strategies in a fully Bayesian decision framework.  Other topics covered are finite and infinite mixtures for Bayesian semi- and non-parametric modeling, large-scale (dynamic/spatial) factor models, Bayesian additive regression trees (BART), Bayesian text modeling and modeling large-scale time-varying covariance matrices.  All classroom examples and implementations as well as projects will be carried out by the open-source statistical software R.

Useful textbooks:

• Gamerman and Lopes (2006) MCMC: Stochastic Simulation for Bayesian Inference, Second Edition. Chapman & Hall/CRC. http://www.dme.ufrj.br/mcmc.
• Gelman, Carlin, Stern, Dunson, Vehtari and Rubin (2020) Bayesian Data Analysis, Third Edition. Chapman & Hall/CRC. http://www.stat.columbia.edu/~gelman/book/BDA3.pdf
• Hoff (2009) A First Course in Bayesian Statistical Methods. Springer.
• Migon, Gamerman and Louzada (2015) Statistical Inference: An Integrated Approach, Second Edition, Chapman & Hall/CRC.

Homework assignments:  HW are to be delivered via a single PDF file to hedibert@gmail.com

1. HW1 – Due at the beginning of class, February 7th, 2023.
2. HW2 – Due at the beginning of class, February 23rd, 2023.
3. HW3 – Due at the beginning of class, March 16th, 2023
4. HW4 – Due at the beginning of class, April 13th, 2023

List of papers for final presentation – Due date: 12pm, May 2nd 2023.

Your final evaluation has two parts AND both need to be turned on May 2nd 2023 at noon: a) A five-page summary of the paper, and b) recorded 15-minute presentation of the paper plus slides.

1. The illusion of the illusion of sparsity – Fava and Lopes (2021), Brazilian Journal of Probability and Statistics, 35(4), 699-720 https://arxiv.org/abs/2009.14296 (Bryan Lietz)
2. A weakly informative default prior distribution for logistic and other regression models – Gelman, Jakulin, Pittau and Zu (2008), Annals of Applied Statistics, 2(4), 1360-1383. https://doi.org/10.1214/08-AOAS191 (Tyler Hoffman)
3. Do forecasts of bankruptcy cause bankruptcy? A machine learning sensitivity analysis – Papakostas, Hahn, Murray, Zhou and Gerakos (2023), Annals of Applied Statistics, 17(1), 711-739.
   https://doi.org/10.1214/22-AOAS1648 (Yang Ba)
4. Forecasting with many predictors using Bayesian additive regression trees – Pruser (2019), Journal of Forecasting, Volume38, Issue7, November 2019, Pages 621-631. https://doi.org/10.1002/for.2587 (Mina Jiang)

TEACHING MATERIAL

Bayesian ingredients

Old homework assignments (spring 2022): HW1 (Solution) + HW2 (Solution) + HW3 (Derivations + R code) + HW4 (Solution)

Pool of papers from final presentation (spring 2022)

Course: APRENDIZAGEM BAYESIANA  – Mestrado Profissional em Economia (MPE)
Professor: Hedibert Freitas Lopes – www.hedibert.org

Monitoria: Henrique Bolfarine

Syllabus: Baixe aqui

Homework assignments

1. HW1: Use the your own y and X  in the attached R code (Due: 7:30pm, Monday, March 9th, 2020)
2. HW2: PCA e FA para dados de consumo de energia em 7 estados brasileiros

Course notes (+ R code & references)

• Bayesian ingredients  
  • Chapter 1 of “Bayes’ Rule: A tutorial Introduction to Bayesian Analysis (Slide)
  • Tiago Mendonca’s shiny for the physicists example
• Bayesian computation
  • iid Bernoulli or logit regression or probit regression? (Rmarkdown)
  • Learning the number of degree of freedom of a Student’s t  (Rmarkdown)
  • Bayesian regression with the normal-gamma (NG) prior
• Bayesian linear regression
  • Motorcycle example
  • A few R packages for Bayesian inference in linear models
  • Boston housing data: ML and Bayesian inference
  • Stan/rstan: Michael Betancourt’s Bayes Sparse Regression

• Bayesian classification via logistic regression
  • Marketing campaigns of a Portuguese banking institution
• Other important modeling structures
  • Factor models (Additional material)
    • Indice de Desenvolvimento Humano
    • US market (data)
    • Bayesian factor models
    • Reducao de ruido com PCA e FA
  • Finite mixture of distributions
• Treed models
  • Classification and regression trees (CART)
    • CART step-by-step for the Boston housing data
    • CART for the motorcycle data
    • ROC and AUC
  • Bayesian CART
  • Bootstrap aggregating  (bagging)
    • Ilustracao do bootstrap
  • Boosting (weak/stronger learners)
  • Random forests
    • Predicao da covid-19 com dados publicos do Hospital Israelita Albert Einstein/SP
  • Bayesian additive regression trees (BART)
• Latent Dirichlet Allocation (LDA)
  • Twitter + BBC + Dickens, Wells, Verne or Austen? + The cat in the hat by Dr. Seuss
• Neural Networks
• Data Science, Marketing and Business

Additional supporting material

• Stan/rstan for posterior inference: Hamiltonian MC (HMC) methods – by Hedibert Lopes
• MC and MCMC: Key References – by Hedibert Lopes
• Mathematics for Machine Learning – by Deisenroth, Faisal and Ong
• Conceitos e analises estatisticas com R e JASP – by  Luis Anunciacao

Course: ECONOMETRICS III 2018 – Doctoral Program in Business Economics
Professor: Hedibert Freitas Lopes – www.hedibert.org

We have 26 1.5h lectures between April 17th and July 12

Objective

The main goal of the course is to make the student familiar with and able to implement univariate and multivariate time series models by using both frequentist and Bayesian approaches. All classroom examples and implementations as well as projects will be carried out by the open-source statistical software R.

Course description

Brief review of frequentist inference followed by the introduction of key ingredients of Bayesian inference, model selection and criticism. An introduction to the main Monte Carlo methods for Bayesian inference: MC integration, resampling, MCMC and sequential MC. Univariate time series models, including AR(F)IMA models, state-space models, Markov switching models, GARCH and stochastic volatility models. Multivariate time series models, including Bayesian VARs and factor-augmented VARs, dynamic factor models, time-varying covariance models.

Course outline

• PART I: Basic univariate time series models
  • AR, MA and ARMA models
  • Unit-root nonstationarity and long-memory processes
  • Seasonal models
• PART II: Basic Bayes
  • Bayesian ingredients
  • Monte Carlo methods
  • Markov chain Monte Carlo methods
• PART III: More univariate time series
  • ARCH/GARCH models
  • EGARCH, GARCH-M, TGARCH
  • Bayesian GARCH
  • Bayesian inference in the local level model
  • Dynamic models
  • Stochastic volatility models
• PART IV: Multivariate time series models
  • Vector autoregressive models
  • Large BVAR, FAVAR, TVP-BVAR & BFAVAR models
  • Factor models
  • Time-varying covariance models

Bibliography

• Gamerman and Lopes (2006) MCMC: Stochastic Simulation for Bayesian Inference, Second Edition. Chapman & Hall/CRC. http://www.dme.ufrj.br/mcmc
• Tsay (2010) Analysis of Financial Time Series, Third Edition. Wiley-Interscience, Probability and Statistics. http://faculty.chicagobooth.edu/ruey.tsay/teaching/fts3
• Tsay (2014) Multivariate Time Series Analysis with R and Financial Applications. Wiley. http://faculty.chicagobooth.edu/ruey.tsay/teaching/mtsbk
• Shumway and Stoffer (2011) Time Series Analysis and Its Applications with R Examples, Third Edition. Springer.  http://www.stat.pitt.edu/stoffer/tsa4

Teaching assistant: William Rojas (williamrojas1212 at gmail dot com)

Office hours: Wednesdays, from 8am to 9:30am, room 201.  Except on May 9th, room Mario Haberfeld.

Homework assignments

1. HW1: Due date: April 19th, 10:30am – Read chapter one and chapter two (Sections 2.1 to 2.5) of Tsay (2010) and turn in exercises 1.1 to 1.5  (pages 25 to 28) and exercises 2.2 to 2.4 (pages 104 to 105).  You can work individually or in pairs.
2. HW2: Due date: May 17th, 10:30am – Turn in exercises 3.1 and 3.2 (pages 70-71) of Hamilton’s (1994) Time Series Analysis and exercises 2.7, 2.8 and 2.9 (pages 105 to 106) of Tsay’s (2010) Analysis of Financial Time Series.  You can work individually or in pairs.
3. HW3: Due date: May 29th, 10:30am – MC Integration, SIR, Gibbs sampler
4. HW4: Due date: July 5th, 10:30am – VAR and BVAR (problems 2.4, 2.5 and 2.6)

EXAMPLES DEVELOPED IN CLASS 

• Class of April 19th, 2018: ACF of white noise and random walk processes
• Class of May 3rd, 2018: AR(3) simulation exercise(R markdown code)
• Class of May 8th, 2018: Sampling distribution of the Dickey-Fuller ratio
• Class of May 11th, 2018: SARIMA for unemployment rate in Sao Paulo (R code)
• Class of May 15th, 2018: Sequential Bayesian learning
• Class of May 17th, 2018: Monte Carlo integration/simulation
• Class of May 22nd, 2018: Gaussian AR(2) model with conditionally conjugate priors: Gibbs Sampler
• Class of May 24th, 2018: Bayesian linear regression +  Binomial model and mixture of betas prior: comparing SIR and Metropolis-Hastings schemes
• Class of May 25th, 2018: Bayesian CAPM
• Class of May 29th, 2018: Gaussian and non-Gaussian GARCH models + Rmarkdown  + Petrobras
• Class of June 5th, 2018:  Modeling time-varying variances via stochastic volatility (SV) models
• Take home exam
• Class of June 14th, 2018: Hidden Markov model: forward filtering, backward sampling (Rmd code)
• Class of June 19th, 2018: Papers for final presentations
• Class of June 21th, 2019: VAR homework (Due date: July 5th 10:30 am)

TEACHING MATERIAL

PART I: Basic univariate time series

• Autoregressive (AR) models and moving average (MA) models (HTML output)
• Unit-root nonstationarity and long-memory processes  (HTML output)
• Seasonal models
  • SARIMA in R – Brazilian industrial production
  • R code for airline data – ARIMA(0,1,1)(0,1,1)[12] (R markdown html output)

PART II: Basic Bayes

• Bayesian ingredients
• Monte Carlo (MC) methods
• Markov chain Monte Carlo (MCMC) methods

PART III: More univariate time series

1. ARCH/GARCH models
   • Petrobras: a first look + R code + Petrobras data
2. EGARCH, GARCH-M, TGARCH
3. Bayesian GARCH
4. Dynamic models (aka state-space models) and stochastic volatility (SV) models
   • R package bsts (R code)
   • More on the bsts package (R code)
   • Petris & Petroni (2011) State space models in R
   • Tusell (2011) Kalman filtering in R
   • Journal of Statistical Software (2011) – Volume 41 – Statistical Software for State Space Methods
   • Local level model with conditionally gaussian errors

PART IV: Multivariate time series

1. Vector autoregressive models
   • R code for VAR, Chapter 2 of Tsay MTS book + packages
2. Large BVAR, FAVAR, TVP-BVAR & BFAVAR
3. Factor models (Standard factor analysis, Spatial dynamic factors, Factor stochastic volatility)
4. Time-varying covariance models
   • Efficient Bayesian inference for multivariate FSV models (Kastner, Fruewirth-Schnatter & Lopes)
   • Factor stochastic volatility with time varying loadings and Markov switching regimes (Lopes & Carvalho)
   • Cholesky realized SV models  (Shirota, Omori, Lopes & Piao)
   • Bayesian inference for stochastic volatility modeling (Lopes & Polson)
   • A Review of Stochastic Volatility: univariate and multivariate models (Platanioti, McCoy & Stephens)

Additional material on Bayesian time series

1. Bayesian Statistics (a very brief introduction) – Ken Rice, April, 2014
2. Lopes and Salazar (2006) Bayesian model uncertainty in smooth transition autoregressions, Journal of Time Series Analysis, 27, 99-117.
3. Huerta and Lopes (2000) Bayesian forecasting and inference in latent structure for the Brazilian industrial production index, Brazilian Review of Econometrics, 20, 1-26.
4. Kleibergen and Hoek (2000) Bayesian Analysis of ARMA Models. Tinbergen Institute Discussion Paper.
5. Marriott, Ravishanker, Gelfand and Pai (1995) Bayesian Analysis of ARMA Processes: Complete Sampling Based Inference under Exact Likelihoods. Bayesian Statistics and Econometrics: Essays in honor of Arnold Zellner. Berry, Chaloner and Geweke, eds., John Wiley & sons, 241-254.

R stuff

Radford Neal’s 13 lectures about R

McLeod, Yu and Mahdi’s (2012) Time Series Analysis with R

Course: BAYESIAN ECONOMETRICS 2017 – Doctoral Program in Business Economics
Professor: Hedibert Freitas Lopes – www.hedibert.org

Syllabus

Objective

The end of the course goal is to allow the student to critically decide between a Bayesian, a frequentist or Bayesian-frequentist compromise when facing real world problems in the fields of micro-econometrics, macro-econometrics, marketing and finance.  With this end in mind, we will visit well known Bayesian issues, such as prior specification and model comparison and model averaging, but also study regularization, “small n, large p” issues, Bayesian statistical learning (additive regression trees) and large-scale factor models.

Course description

Basic ingredients: prior, posterior, and predictive distributions, sequential Bayes, conjugate analysis, exchangeability, principles of data reduction and decision theory.  Model criticism: Bayes factor, computing marginal likelihoods, Savage-Dickey ratio, reversible jump MCMC, Bayesian model averaging and deviance information criterion.  Modern computation via (Markov chain) Monte Carlo methods: Monte Carlo integration, sampling-importance resampling, Gibbs sampler, Metropolis-Hastings algorithms.  Mixture models, Hierarchical models, Bayesian regularization, Instrumental variables modeling, Large-scale (sparse) factor modeling, Bayesian additive regression trees (BART) and related topics, Dynamic models, Sequential Monte Carlo algorithms, Bayesian methods in microeconometrics, macroeconometrics, marketing and finance

Course notes (+ R code & references)

• Bayesian ingredients, Monte Carlo and Markov chain Monte Carlo methods
  • Exchangeability + Principles of data reduction + More on estimators + Decision theory
  • Monte Carlo integration for a (very, very) simple normal-normal example
  • Two models & three priors – a Bayesian exercise via sampling importance resampling (R code)
    
  • Gibbs sampler for AR(1) model with one changing point on the persistence coefficient (changing in the intercept)
  • Posterior inference for mean and variance of i.i.d. data from student’s t with known degrees of freedom (R code)
• Bayesian model criticism
• General linear and hierarchical models
• Limited dependent variable models
• Factor and spatial models
• Mixture models
• Bayesian instrumental variables
• Dynamic and stochastic volatility models
• Sequential Monte Carlo methods
  • SMC: bootstrap filter
  • SMC: bootstrap filter, auxiliary particle filter, optimal particle filter
  • SMC with parameter learning
  • Bootstrap filter for stochastic volatility

Miscellaneous

• The R project
• Hedibert Lopes’ Bayesian R packages for Econometrics
• David Hiebeler’s R & Matlab
• R for Matlab users
• David Li’s performance comparison among: R, Matlab, SAS, SPSS, Excel
• WinBUGS project
• Gamerman and Lopes’ website

Course: ECONOMETRICS III 2017 – Doctoral Program in Business Economics
Professor: Hedibert Freitas Lopes – www.hedibert.org

Syllabus

Objective

The main goal of the course is to make the student familiar with and able to implement univariate and multivariate time series models by using both frequentist and Bayesian approaches. All classroom examples and implementations as well as projects will be carried out by the open-source statistical software R.

Course description

Brief review of frequentist inference followed by the introduction of key ingredients of Bayesian inference, model selection and criticism. An introduction to the main Monte Carlo methods for Bayesian inference: MC integration, resampling, MCMC and sequential MC. Univariate time series models, including AR(F)IMA models, state-space models, Markov switching models, GARCH and stochastic volatility models. Multivariate time series models, including Bayesian VARs and factor-augmented VARs, dynamic factor models, time-varying covariance models.

Bibliography

• Gamerman and Lopes (2006) MCMC: Stochastic Simulation for Bayesian Inference, Second Edition. Chapman & Hall/CRC. http://www.dme.ufrj.br/mcmc
• Tsay (2010) Analysis of Financial Time Series, Third Edition. Wiley-Interscience, Probability and Statistics. http://faculty.chicagobooth.edu/ruey.tsay/teaching/fts3
• Tsay (2014) Multivariate Time Series Analysis with R and Financial Applications. Wiley. http://faculty.chicagobooth.edu/ruey.tsay/teaching/mtsbk
• Shumway and Stoffer (2011) Time Series Analysis and Its Applications with R Examples, Third Edition. Springer.  http://www.stat.pitt.edu/stoffer/tsa4

Teaching assistant: Paloma Uribe (paloma dot uribe at gmail dot com)

Office hours: Thursdays from 1pm to 2pm (Room 604)

Day-to-day announcements

Homework assignments & Exams

• homework 1
  • Solution to problem 2b and 2c  (1.4 secs on a 2012 MacBook Pro, 2.6 GHz Intel Core i7, 8 GB 1600 MHz DDR3)
  • Additional MC exercise I decided to do on my own (2.7 secs)
  • Solution to problem 2d (90 secs)
• midterm exam (solution)
• homework 2 (dataset)

April, 18th to 24th 2017:

• R code for the AR(2) example worked in class
• R for Shumway and Stoffer’s chapter 1
• More R code for the AR(1) and AR(2) processes  (Slides)
• R markdown script (run via Rstudio) (PDF output or HTML output)
• Please, read/study the following material – Tsay (2010): Chapter 1 – read/study it only if you need a quick review of basic stats. Chapter 2 – read particularly sections 2.1 to 2.4 on stationarity, autocorrelation function, white noise and simple AR models.  Try working on problem 2.2.  Shumway and Stoffer (2011): Chapter 1 – read examples 1.1 to 1.12 and section 1.5 (stationary time series) and try working on problems 1.1 to 1.6 and 1.8.  Chapter 3 – read examples 3.1, 3.2, 3.7, 3.8, 3.9 and 3.12 and try working on problems  3.5 and 3.6.

April, 25th to 30th 2017:

• Monte Carlo exercise: studying the sampling behavior of the t test under unit root

May, 1st to May 7th 2017:

• Bayesian inference for the Gaussian AR(2) model (R code)

May 8th to May 15th 2017:

This week we started talking about two relatively simple and well-known time series models with time-varying variances: ARCH(1) and GARCH(1,1) models.  We also considered a standard Bayesian approach to posterior inference regarding the GARCH(1,1) model.  Chapter 3 of Tsay’s (2010) book is good enough as an introduction to the subject, particularly sections 3.4 (ARCH model) and 3.5 (GARCH model).  A few exercises from his book are worth trying to fix most the ideas.  I recommend the following exercises: 3.1, 3.2, 3.3, 3.4, 3.5 and 3.6.

• R code for computing pi via rejection sampling – our first Monte Carlo sampling scheme
• Monte Carlo integration for a (very, very) simple normal-normal example
• Gibbs sampler for AR(1) model with one changing point on the persistence coefficient (changing in the intercept)

May, 16th to May 23rd 2017:

• Brazilian monthly production of cement (January 2002 to February 2017)
• AR(1) plus noise: Figure 1 +Figure 2 + R code
• AR(1) plus noise – Kalman filter and smoother: Figures + R code
• AR(1) plus noise – Bayesian inference via MCMC/FFBS: Figures + R code
• AR(1) plus noise – Modeling Alcoa realized volatilities via 1st order DLM: Data + R code

May, 24th to June 5th 2017:

• AR(1) plus noise – Comparing block move (FFBS) with single move MCMC schemes:  HTML

June 6th to June 15th 2017:

• Linear regression with Markov switching intercept – R code + Figures

COURSE SLIDES AND TEACHING MATERIAL

PART I: Basic univariate time series

• Autoregressive (AR) models and moving average (MA) models (HTML output)
• Unit-root nonstationarity and long-memory processes  (HTML output)
• Seasonal models
  • SARIMA in R – Brazilian industrial production
  • R code for airline data – ARIMA(0,1,1)(0,1,1)[12] (R markdown html output)

PART II: Basic Bayes

• Bayesian ingredients
• Monte Carlo (MC) methods
• Markov chain Monte Carlo (MCMC) methods

PART III: More univariate time series

1. ARCH/GARCH models
   • Petrobras: a first look + R code + Petrobras data
2. EGARCH, GARCH-M, TGARCH
3. Bayesian GARCH
4. Dynamic models (aka state-space models) and stochastic volatility (SV) models
   • R package bsts (R code)
   • More on the bsts package (R code)
   • Petris & Petroni (2011) State space models in R
   • Tusell (2011) Kalman filtering in R
   • Journal of Statistical Software (2011) – Volume 41 – Statistical Software for State Space Methods
   • Local level model with conditionally gaussian errors

PART IV: Multivariate time series

1. Vector autoregressive models
   • R code for VAR, Chapter 2 of Tsay MTS book + packages
2. Large BVAR, FAVAR, TVP-BVAR & BFAVAR
3. Factor models (Standard factor analysis, Spatial dynamic factors, Factor stochastic volatility)
4. Time-varying covariance models
   • Parsimony inducing priors for large scale state-space models (Lopes, McCulloch & Tsay)
   • Efficient Bayesian inference for multivariate FSV models (Kastner, Fruewirth-Schnatter & Lopes)
   • Factor stochastic volatility with time varying loadings and Markov switching regimes (Lopes & Carvalho)
   • Cholesky realized SV models  (Shirota, Omori, Lopes & Piao)
   • Scalable inference for a full multivarite SV model (Dellaportas, Plataniotis & Titsias)
   • Bayesian inference for stochastic volatility modeling (Lopes & Polson)
   • A Review of Stochastic Volatility: univariate and multivariate models (Platanioti, McCoy & Stephens)
   • A Tractable State-Space Model for Symmetric Positive-Definite Matrices (Windle & Carvalho)
   • Autoregressive Models for Variance Matrices: Stationary Inverse Wishart Processes (Fox & West)
   • Multivariate stochastic volatility with Bayesian dynamic linear models (Triantafyllopoulos)
   • On Singular Wishart and Singular Multivariate Beta Distributions (Uhlig)
   • Bayesian Vector Autoregressions with Stochastic Volatility (Uhlig)
   • Multivariate Stochastic Volatility via Wishart Processes (Philipov & Glickman)
   • Factor multivariate stochastic volatility via Wishart processes (Philipov & Glickman)
   • The Structure of Dynamic Correlations in Multivariate Stochastic Volatility Models (Asai and McAleer)
   • Multivariate GARCH Models: a Survey (Bauwens, Laurent and Rombouts)

Additional material

1. Bayesian Statistics (a very brief introduction) – Ken Rice, April, 2014
2. Lopes and Salazar (2006) Bayesian model uncertainty in smooth transition autoregressions, Journal of Time Series Analysis, 27, 99-117.
3. Huerta and Lopes (2000) Bayesian forecasting and inference in latent structure for the Brazilian industrial production index, Brazilian Review of Econometrics, 20, 1-26.
4. Kleibergen and Hoek (2000) Bayesian Analysis of ARMA Models. Tinbergen Institute Discussion Paper.
5. Marriott, Ravishanker, Gelfand and Pai (1995) Bayesian Analysis of ARMA Processes: Complete Sampling Based Inference under Exact Likelihoods. Bayesian Statistics and Econometrics: Essays in honor of Arnold Zellner. Berry, Chaloner and Geweke, eds., John Wiley & sons, 241-254.

R stuff

Introduction to R (by Paloma Uribe, in Portuguese) 

Radford Neal’s 13 lectures about R

McLeod, Yu and Mahdi’s (2012) Time Series Analysis with R

Course: ECONOMETRICS III 2016 – Doctoral Program in Business Economics
Professor: Hedibert Freitas Lopes – www.hedibert.org

Syllabus

Objective

The main goal of the course is to make the student familiar with and able to implement univariate and multivariate time series models by using both frequentist and Bayesian approaches. All classroom examples and implementations as well as projects will be carried out by the open-source statistical software R.

Course description

Brief review of frequentist inference followed by the introduction of key ingredients of Bayesian inference, model selection and criticism. An introduction to the main Monte Carlo methods for Bayesian inference: MC integration, resampling, MCMC and sequential MC. Univariate time series models, including AR(F)IMA models, state-space models, Markov switching models, GARCH and stochastic volatility models. Multivariate time series models, including Bayesian VARs and factor-augmented VARs, dynamic factor models, time-varying covariance models.

Bibliography

• Gamerman and Lopes (2006) MCMC: Stochastic Simulation for Bayesian Inference. Chapman & Hall/CRC. http://www.dme.ufrj.br/mcmc
• Tsay (2010) Analysis of Financial Time Series, Third Edition. Wiley-Interscience, Probability and Statistics. http://faculty.chicagobooth.edu/ruey.tsay/teaching/fts3
• Tsay (2014) Multivariate Time Series Analysis with R and Financial Applications. Wiley. http://faculty.chicagobooth.edu/ruey.tsay/teaching/mtsbk

Course material

1. Bayesian ingredients
   • GL2006: Examples 2.1, 2.4, 2.5, section 2.3.2, problems 2.1, 2.2, 2.5, 2.8a, 2.10 and 2.11.
2. Monte Carlo methods
3. Markov chain Monte Carlo methods
4. Autoregressive (AR) models and moving average (MA) models
   • Tsay (2010): Sections 2.1-2.6, sections 2.7 and 2.11, and sections 3.1-3.8
5. Unit-root nonstationarity and long-memory processes
6. Seasonal models
7. ARCH/GARCH models (EGARCH, GARCH-M, TGARCH) (Bayesian GARCH)
8. Dynamic models (aka state-space models) and stochastic volatility (SV) models
   • R package bsts (R code)
   • Petris & Petroni (2011) State space models in R
   • Tusell (2011) Kalman filtering in R
   • Journal of Statistical Software (2011) – Volume 41 – Statistical Software for State Space Methods
   • Local level model with conditionally gaussian errors
9. Homework (due June 1st)  (solution)
10. Vector autoregressive models
    • R code for VAR, Chapter 2 of Tsay MTS book
    • Packages
11. Large BVAR, FAVAR, TVP-BVAR & BFAVAR
12. Sequential Monte Carlo methods
13. Factor models (Standard factor analysis, Spatial dynamic factors, Factor stochastic volatility)
14. Time-varying covariance models
    • Parsimony inducing priors for large scale state-space models (Lopes, McCulloch & Tsay)
    • Efficient Bayesian inference for multivariate FSV models (Kastner, Fruewirth-Schnatter & Lopes)
    • Factor stochastic volatility with time varying loadings and Markov switching regimes (Lopes & Carvalho)
    • Cholesky realized SV models  (Shirota, Omori, Lopes & Piao)
    • Scalable inference for a full multivarite SV model (Dellaportas, Plataniotis & Titsias)
    • Bayesian inference for stochastic volatility modeling (Lopes & Polson)
    • A Review of Stochastic Volatility: univariate and multivariate models (Platanioti, McCoy & Stephens)
    • A Tractable State-Space Model for Symmetric Positive-Definite Matrices (Windle & Carvalho)
    • Autoregressive Models for Variance Matrices: Stationary Inverse Wishart Processes (Fox & West)
    • Multivariate stochastic volatility with Bayesian dynamic linear models (Triantafyllopoulos)
    • On Singular Wishart and Singular Multivariate Beta Distributions (Uhlig)
    • Bayesian Vector Autoregressions with Stochastic Volatility (Uhlig)
    • Multivariate Stochastic Volatility via Wishart Processes (Philipov & Glickman)
    • Factor multivariate stochastic volatility via Wishart processes (Philipov & Glickman)
    • The Structure of Dynamic Correlations in Multivariate Stochastic Volatility Models (Asai and McAleer)
    • Multivariate GARCH Models: a Survey (Bauwens, Laurent and Rombouts)

Additional material

1. Bayesian Statistics (a very brief introduction) – Ken Rice, April, 2014
2. Lopes and Salazar (2006) Bayesian model uncertainty in smooth transition autoregressions, Journal of Time Series Analysis, 27, 99-117.
3. Huerta and Lopes (2000) Bayesian forecasting and inference in latent structure for the Brazilian industrial production index, Brazilian Review of Econometrics, 20, 1-26.
4. Kleibergen and Hoek (2000) Bayesian Analysis of ARMA Models. Tinbergen Institute Discussion Paper.
5. Marriott, Ravishanker, Gelfand and Pai (1995) Bayesian Analysis of ARMA Processes: Complete Sampling Based Inference under Exact Likelihoods. Bayesian Statistics and Econometrics: Essays in honor of Arnold Zellner. Berry, Chaloner and Geweke, eds., John Wiley & sons, 241-254.

Disciplina: ECONOMETRIA – Turma 4ECO (Economia)
Período Letivo: 2016/2
Professor: Hedibert Freitas Lopes – www.hedibert.org
Monitora: Paloma Vaissman Uribe – PalomaVU@insper.edu.br

Programa de Ensino 

Conteudo das aulas

1. Apresentacao do curso
   • wage.R + wage.txt + wage.csv
   • violencia.R +  violencia.csv (Mapa da Violencia 2015)
2. Regressao Linear Simples – Parte 1: Minimos Quadrados Ordinarios & R2
   • VendedoresRH.xls
   • Mais sobre minimos quadrados ordinarios
   • Modelando salario via anos de experiencia, posicao e sexo (Codigo R)  (Dataset)
   • House prices: preco, tamanho, quarto, banheiros, etc (Codigo R)
   • Mais sobre R2: explicando GPA (Codigo R)
   • Gauss and the invention of least squares (by Stigler, 1981)
   • The discovery of the method of least squares (by Plackett, 1972)
3. Regressao Linear Simples – Parte 2: formas funcionais
   • Mais sobre formas funcionais
   • Modelos log-log, nivel-log, log-nivel, nivel-nivel para dados HOUSE2 (Codigo R)
   • Modelos log-log, nivel-log, log-nivel, nivel-nivel para dados TESTESCORE (Codigo R)
4. Regressao Linear Simples – Parte 3: Suposicoes, propriedades e teste-t
   • Comportamento do estimador da inclinacao
   • caliescom.txt + caliescom.R
   • height x weight, nba-wnba players + nba.csv + wnba.csv +  Codigo R
   • Records dos 100 metros rasos + dados + Codigo R
5. Analise de residuos
   • houseprices-regressaomultipla + Codigo R
6. Regressao linear multipla
   • temco.txt: conjunto de dados utilizado nas notas de aula sobre regressao linear simples.
   • Bank wages: Mais sobre R2 e R2 ajustado para 15 modelos (Codigo R)  (Dataset)
   • Data on monthly earnings, education, several demographic variables, and IQ scores for 935 men in 1980. (Dataset)  (Codigo R)
   • Data on 4,137 US college students.(Dataset)  (Codigo R)
   • Tutorial em R – Parte I (Datasets: salario.txt & salario.csv)
   • Tutorial em R – Parte II :(Dataset: houseprices.txt)
   • Precos de imoveis via dummies – houseprices.txt
   • Precos de imoveis via dummies & interacoes – houseprices.txt  – Codigo R
   • Salarios de um banco americano – bankwages.txt – Codigo R
   • Precos de sucos de laranja – orangejuice-chicagoarea.csv – Codigo R
   • Precos de Toyota Corollas usados – toyotacorolla.csv – Codigo R
   • Peso vs altura, idade, sexo – peso-altura-idade-sexo-criancas.txt – Codigo R
   • Monthly earnings, education, IQ, etc for 935 men in 1980 – wage2-wooldridge.txt – Codigo R
7. Regressao linear multipla – Vies de omissao de variable
8. Regressao linear multipla – Teste F Parcial
9. Heteroscedasticidade
   • Exercicio de simulacao
   • Worked example – Codigo R
10. Endogeneidade: variaveis instrumentais (worked examples)
    • Algumas ilustracoes
11. Endogeneidade: estimacao
12. Endogeneidade: tests
    • Return to education (women) – Codigo R
13. Basic time series  – Codigo R
    • Algumas series temporais (ipi.csv – pbr.csv)

Listas de exercicios

Lista 1 (Regressao linear simples): Wooldridge – 2.2, 2.3, 2.4, 2.5, 2.7, 2.9, 2.11

Lista 2 (Regressao linear multipla): Wooldridge – 3.3, 3.4, 3.5, 3.7, 3.9, 4.2, 4.3, 4.6, 4.9, 4.11

Lista 3 (Heteroscedasticidade): Wooldridge – 8.1, 8.2, 8.3, 8.4, 8.5 + refazer os exemplos 8.1 (pg 250-251), 8.2 (pg 252), 8.4 (pg 257) e 8.7 (pg 268-269)

Lista 4 (Endogeneidade): Wooldridge – 15.1, 15.2, 15.3, 15.7, 15.8 e 15.10 + refazer os exemplos 5.1 (pg 476-477), 5.2 (pg 477-478), 5.3 (pg 480-481) e 5.4 (pg 484-486).

Lista 5 (Series temporais): Wooldridge – Refazer os exemplos 10.1, 10.2, 10.3, 10.4, 10.7, 10.9, 11.3, 11.4, 11.5, 11.6 e 11.7 + problemas 11.1, 11.2, 11.3, 11.4 e 11.5

Atividade 2

Dados por escola do ENEM2015 – Analise exploratoria dos dados  –  Codigo R

Econometria em R

1. Tutorial de R: aula 1
2. Tutorial de R: aula 2
3. Tutorial de R: regressao
4. Tutorial de R: R2
5. Introducao ao uso do R (Paloma Uribe)
6. Using R for Introductory Econometrics (Florian Heiss)
7. Econometric and time series modeling using R (Cribari-Neto)
8. Introduction to programmingEconometrics with R (Bruno Rodrigues)
9. Econometrics in R: Past, Present, and Future (Achim Zeileis & Roger Koenker)
10. CRAN Task View: Econometrics (Achim Zeileis)
11. R-Econometrics – Learn R for applied economics in a comprehensive way

Econometria em outras linguagens/pacotes

1. PYTHON: Introductory Econometrics – Jeffrey M. Wooldridge: Capítulos 2 ao 8 usando PYTHON
2. Kevin Sheppard’s Python for Econometrics
3. STATA: Introductory Econometrics – Jeffrey M. Wooldridge: Capítulos 2 ao 18 usando STATA
4. Statistical Analysis in R, MATLAB, SAS, STATA and SPSS

Mais conjuntos de dados

• • • VendedoresRH.xls
    • salario.txt: Salario vs posicao, anos de experiencia e sexo
    • retornos-2014.csv: Retornos de Ambev, Vale, Petrobras, JBS, Natura, Gafisa, Lojas Americanas & Ibovespa
    • vs posicao, anos de experiencia e sexo
    • bankwages.txt: Salario vs salario inicial, educacao, sexo, minoria e categoria
    • wage2-wooldridge.txt: Monthly earnings, education, IQ, etc for 935 men in 1980
    • gpa2-wooldridge.txt: Data on 4,137 US college students
    • houseprices.txt: House prices vs offers, sqft, brick, bedrooms, bathrooms and neighborhood
    • orangejuice-chicagoarea.csv: Weekly sales of 64oz orange juice containers in the Chicago area
    • toyotacorolla.csv: Sales prices and vehicle characteristics of 1436 used Toyota Corollas
    • peso-altura-idade-sexo-criancas.txt: peso vs altura, idade e sexo
    • gpa2.txt: same as gpa2-wooldridge.txt
    • temco.txt
    • temcoprod.txt
    • hprice1.txt: dados sobre 88 residencias.
    • dadosmunicipais.csv: dados de 5848 municipios brasileiros (analfabetismo, renda, desigualdade).
    • caliescom.xls: Performance media de 420 escolas em um teste padronizado.
    • ceosal2.xls: Data on 177 chief executive officers.
    • florida2000.xls: Florida 2000 presidential vote as of 5pm ET Saturday, Nov. 11, 2000 (after recount).
    • temcoprod.xls: Caracteristicas de funcionarios de uma firma.
    • simulados_producao.xls: Modelagem da produção de firmas de um determinado setor.
    • wage1.csv: Wooldridge’s dataset.
    • smoke.txt: Wooldridge’s dataset.
    • card.csv: Wooldridge’s dataset.
    • mroz.csv: Wooldridge’s dataset.
    • bwght.csv: Wooldridge’s dataset.

Disciplina: ECONOMETRIA – Turma 4ECO (Economia)
Período Letivo: 2016/1
Professor: Hedibert Freitas Lopes – www.hedibert.org
Monitora: Paloma Vaissman Uribe – PalomaVU@insper.edu.br

Programa de Ensino 

Conteudo das aulas

1. Apresentacao do curso
   • wage.R + wage.txt + wage.csv
   • violencia.R +  violencia.csv (Mapa da Violencia 2015)
2. Regressao Linear Simples – Parte 1: Minimos Quadrados Ordinarios & R2
   • VendedoresRH.xls
   • Mais sobre minimos quadrados ordinarios
   • Modelando salario via anos de experiencia, posicao e sexo (Codigo R)  (Dataset)
   • House prices: preco, tamanho, quarto, banheiros, etc (Codigo R)
3. Regressao Linear Simples – Parte 2: formas funcionais
4. Regressao Linear Simples – Parte 3: Suposicoes, propriedades e teste-t
   • caliescom.txt + caliescom.R
   • height-weight-nba.csv + height-weight-wnba.csv +  height-weight.R
5. Analise de residuos
   • houseprices-regressaomultipla.R
6. Atividade 1 – Dados da PNAD 2009 + readme.txt + Analise exploratoria dos dados
   • Madalozzo and Mauriz (2012) Does investing in education reduce the gender wage gap? A Brazilian population study. Population Review, Volume 51, Number 2, pp. 59-84.
   • Bertrand, Kamenica and Pan (2015) Gender Identity and relative income within households.  Quarterly Journal of Economics.
   • USA data: Median usual weekly earnings (second quartile), Employed full time, Wage and salary workers.  U.S. Bureau of Labor Statistics, United States Department of Labor.
   • pnad2009.R: Codigo R para os dados brasileiros.
   • earnings.R: Codigo R para os dados americanos.
7. Regressao linear multipla
   • temco.txt: conjunto de dados utilizado nas notas de aula sobre regressao linear simples.
   • R da aula de 16/03/2016
   • R da aula de 23/03/2016
   • R da aula de 30/03/2016
   • Bank wages: Mais sobre R2 e R2 ajustado para 15 modelos (Codigo R)  (Dataset)
   • Data on monthly earnings, education, several demographic variables, and IQ scores for 935 men in 1980. (Dataset)  (Codigo R)
   • Data on 4,137 US college students.(Dataset)  (Codigo R)
   • Tutorial em R – Parte I (Datasets: salario.txt & salario.csv)
   • Tutorial em R – Parte II :(Dataset: houseprices.txt)
   • Precos de imoveis via dummies – houseprices.txt
   • Precos de imoveis via dummies & interacoes – houseprices.txt  – Codigo R
   • Salarios de um banco americano – bankwages.txt – Codigo R
   • Precos de sucos de laranja – orangejuice-chicagoarea.csv – Codigo R
   • Precos de Toyota Corollas usados – toyotacorolla.csv – Codigo R
   • Peso vs altura, idade, sexo – peso-altura-idade-sexo-criancas.txt – Codigo R
   • Monthly earnings, education, IQ, etc for 935 men in 1980 – wage2-wooldridge.txt – Codigo R
8. Regressao linear multipla – Teste F Parcial + Teste RESET
9. Heteroscedasticidade
   • Exercicio de simulacao
   • Worked example – Codigo R
10. Dados para atividade 2: sleep75.csv
11. Endogeneidade: variaveis instrumentais (worked examples)
12. Endogeneidade: estimacao
13. Endogeneidade: tests
    • Return to education (women) – Codigo R
14. Basic time series  – Codigo R

Listas de exercicios

1. Lista 1
2. Lista 2
3. Lista 3
4. Lista 4

Econometria em R

1. Introducao ao uso do R (Paloma Uribe)
2. Using R for Introductory Econometrics (Florian Heiss)
3. Econometric and time series modeling using R (Cribari-Neto)
4. Introduction to programmingEconometrics with R (Bruno Rodrigues)
5. Econometrics in R: Past, Present, and Future (Achim Zeileis & Roger Koenker)
6. CRAN Task View: Econometrics (Achim Zeileis)
7. R-Econometrics – Learn R for applied economics in a comprehensive way

Econometria em outras linguagens/pacotes

1. PYTHON: Introductory Econometrics – Jeffrey M. Wooldridge: Capítulos 2 ao 8 usando PYTHON
2. Kevin Sheppard’s Python for Econometrics
3. STATA: Introductory Econometrics – Jeffrey M. Wooldridge: Capítulos 2 ao 18 usando STATA
4. Statistical Analysis in R, MATLAB, SAS, STATA and SPSS

Mais conjuntos de dados

• • • VendedoresRH.xls
    • salario.txt: Salario vs posicao, anos de experiencia e sexo
    • retornos-2014.csv: Retornos de Ambev, Vale, Petrobras, JBS, Natura, Gafisa, Lojas Americanas & Ibovespa
    • vs posicao, anos de experiencia e sexo
    • bankwages.txt: Salario vs salario inicial, educacao, sexo, minoria e categoria
    • wage2-wooldridge.txt: Monthly earnings, education, IQ, etc for 935 men in 1980
    • gpa2-wooldridge.txt: Data on 4,137 US college students
    • houseprices.txt: House prices vs offers, sqft, brick, bedrooms, bathrooms and neighborhood
    • orangejuice-chicagoarea.csv: Weekly sales of 64oz orange juice containers in the Chicago area
    • toyotacorolla.csv: Sales prices and vehicle characteristics of 1436 used Toyota Corollas
    • peso-altura-idade-sexo-criancas.txt: peso vs altura, idade e sexo
    • gpa2.txt: same as gpa2-wooldridge.txt
    • temco.txt
    • temcoprod.txt
    • hprice1.txt: dados sobre 88 residencias.
    • dadosmunicipais.csv: dados de 5848 municipios brasileiros (analfabetismo, renda, desigualdade).
    • caliescom.xls: Performance media de 420 escolas em um teste padronizado.
    • ceosal2.xls: Data on 177 chief executive officers.
    • florida2000.xls: Florida 2000 presidential vote as of 5pm ET Saturday, Nov. 11, 2000 (after recount).
    • temcoprod.xls: Caracteristicas de funcionarios de uma firma.
    • simulados_producao.xls: Modelagem da produção de firmas de um determinado setor.
    • wage1.csv: Wooldridge’s dataset.
    • smoke.txt: Wooldridge’s dataset.
    • card.csv: Wooldridge’s dataset.
    • mroz.csv: Wooldridge’s dataset.
    • bwght.csv: Wooldridge’s dataset.

Disciplina: ANALISE MULTIVARIADA 2015 (MPA)

Período Letivo: 2015/1

Professor: Hedibert Freitas Lopes – www.hedibert.org

Monitor: Leandro Augusto Ferreira

Objetivo: O objetivo do curso é apresentar os conceitos e métodos de análise multivariada de dados, aplicando-os a dados reais e interpretando os resultados de forma prática. No curso de análise multivariada são utilizados conceitos de estatística básica e inferência, com ênfase na resolução de problemas reais e interpretação dos resultados.  Na maioria dos estudos, a complexidade dos fenômenos estudados faz com seja necessário coletar informações sobre um conjunto de variáveis. A análise multivariada permite o estudo simultâneo de um conjunto de variáveis, aproveitando a estrutura de correlação existente entre as mesmas.  Nesta disciplina são apresentadas técnicas de análise de dados quantitativos e qualitativos, discutindo aplicações nas áreas de marketing, operações, recursos humanos e finanças.

Programa de Ensino

03/02/2015: Análise Exploratória de Dados Multivariados

• • • Codigo R: descritiva – vinho – ILE-2013-R – bovespa

06/02/2015: Inferencia Multivariada – MANOVA

• • • Codigo R: manova – bebidas

10/02/2015: Análise de Componentes Principais

• • • Codigo R:  bebidas-princomp – mundo – USArrests

24/02/2015: Análise Fatorial – parte 1 + Análise Fatorial – parte 2

• • • Codigo R:  exchange rates – british cohort study – retornos bolsa valores SP

03/03/2015: Regressao logistica + Analise discriminante

• • • Codigo R:  Trator-RL – Arthritis-RL – Trator-LDA

10/03/2015: Correlacao canonica

• • • Codigo R:  performance – psychometrics – carmarks – sales

17/03/2015: Cluster analysis

• • • Codigo R:  ILE-2013-cluster – IDHML-IDHMR

24/03/2015: Trabalho em sala de aula (terminado em casa) 31/03/2015: Correspondence analysis & multidimensional scaling

• • • IDHM brasileiro: Documentacao + Siglas +  Dados +  Codigo R
    • Mais codigo R: correspondenceanalysis + multidimensionalscaling
    • 07/04/2015: Conjoint analysis
    • Mensuração da estrutura de preferência do consumidor (J.O.Siqueira, Tese de mestrado, FEA/USP)

10/04/2015: Structural equation modeling

Conjunto de dados:

• • • descritiva.csv
    • vinho.txt
    • ILE-2013.csv
    • bovespa.csv
    • bebidas.csv
    • mundo.txt
    • exchangerate-monthly.txt
    • BCS-males.txt
    • Ret2012.csv
    • Trator.txt
    • arthritis.txt
    • vendedores.csv
    • carmars.txt

Course: Readings in Statistics and Econometrics 2016: Bayesian Statistical Learning

Professors: Hedibert Freitas Lopes & Paulo Marques

Objective: In this Second Readings in Statistics and Econometrics we will study and discuss, through a series of well established papers, the broad topic of Statistical Learning with an emphasis on its natural Bayesian solutions. The 5 lectures and 8 seminars will take place on Fridays between 10am and 12pm from January 29th to April 8th 2016. Paulo and I will give lectures discussing traditional Statistical Learning techniques, alternated with seminars given by the participants on papers presenting Bayesian counterparts to the techniques discussed in the lectures.

Readings in Statistics and Econometrics 2015: Causality 

Organizer: Hedibert Freitas Lopes

Email: hedibertFL at insper.edu.br

In this First Readings in Statistics and Econometrics we will study and discuss, through a series of well established papers, the broad topic of causality.  The lectures are held at INSPER on Tuesdays, from 7:30am to 9:30am, from September 29th to December 1st, 2015, at classroom Paulo Renato de Souza, 2nd floor.

Annotated bibliography: Here you will find links to textbooks and edited books, special issues, articles with discussion and web material: slides of lectures, discussion of causality, video lectures and more (in chronological order).

Annotated bibliography: Only articles and book chapters (in alphabetical order).

Outline of the lectures

• • • September 29th – Hedibert Lopes – INSPER
      Haavelmo (1943) The statistical implications of a system of simultaneous equations. Econometrica, 11, 1-12.
      slides of the lecture
    • October 6th – Hedibert Lopes – INSPER
      Rubin (1974) Estimating causal effects of treatments in randomized and nonrandomized studies. Journal of Educational Psychology, 56, 688-701.
      slides of the lecture
    • October 13th – Andre Yoshizumi, IME/USP
      Holland (1986) Statistics and causal inference (with discussion). JASA, 81, 945-970.
      slides of the lecture
    • October 20th – Paloma Uribe, IME/USP
      Pearl (1995) Causal diagrams for empirical research (with discussion). Biometrika, 82, 669-710.
      slides of the lecture + slides of Joao M.P.De Mello’s talk
    • November 3rd – Sergio Firpo, EESP/FGV
      Angrist, Imbens and Rubin (1996) Identication of causal effects using instrumental variables (with discussion). JASA, 91, 444-472.
      slides of the lecture
    • November 10th – Julio Trecenti, IME/USP
      Dawid (2000) Causal inference without counterfactuals (with discussion). JASA, 95, 407-424.
      slides of the lecture
    • November 24th – Manasses Nobrega, UFABC
      Vansteelandt and Goetghebeur (2003) Causal inference with generalized structural mean
      models. JRSS-B, 65, 817-835.
    • December 1st – Hedibert Lopes – INSPER
      Heckman and Pinto (2015) Causal analysis after Haavelmo. Econometric Theory, 31,115-151.
      slides of the lecture

Books & special issues

• • • Journal of Econometrics (1988), Volume 39, Issues 1-2
    • Spirtes, Glymour and Scheines (2001) Causation, Prediction, and Search (2nd edition)
      
    • Gelman and Meng (2004) Applied Bayesian Modeling and Causal Inference from Incomplete-Data Perspectives
    • Dawid (2007) Fundamentals of Statistical Causality
    • Morgan and Winship (2007) Counterfactuals and Causal Inference: Methods and Principles for Social Research (2nd edition)
    • Angrist and Pischke (2008) Mostly Harmless Econometrics: An Empiricist’s Companion
    • Pearl (2009) Causality: Models, Reasoning and Inference (2nd Edition)
    • Schroeder (2010) Accounting and Causal Effects: Econometric Challenges
    • Berzuini, Dawid and Bernardinelli (2012) {\em Causality: Statistical Perspectives and Applications
    • Morgan (2013) Handbook of Causal Analysis for Social Research
    • Imbens and Rubin (2015) Causal Inference for Statistics, Social, and Biomedical Sciences: An Introduction
    • Hernan and Robins (2015) Causal Inference
    • Econometric Theory (2015), Volume 31, Issue 01

Articles with discussion

• • • Holland (1986) Statistics and causal inference. JASA, 81, 945-970.
    • Pearl (1995) Causal diagrams for empirical research. Biometrika, 82, 669-710.
    • Angrist, Imbens and Rubin (1996) Identification of causal effects using IVs. JASA, 91, 444-472.
    • Dawid (2000) Causal inference without counterfactuals. JASA, 95, 407-424.
    • Heckman (2005) The scientific model of causality. Sociological Methodology, 35, 1-150.

From the web

• • • Cosma Shalizi’s page
    • Andrew Gelman’s blog:
      • The randomized experiment as gold standard?
      • Resolving disputes between J. Pearl and D. Rubin on causal inference
      • Philip Dawid’s explication of Pearl’s model, and two ways of thinking about nonrandom sampling
      • More on Pearl/Rubin, this time focusing on a couple of points
      • The Roy causal model?
      • Why ask why? Forward causal inference and reverse causal questions
    • Judea Pearl’s home
    • Journal of Causal Inference
    • Drawing Causal Inference from Big Data (Sackler YouTube Channel)
    • Athey and Imbens on Machine Learning
    • Imbens and Wooldrige: Whats New in Econometrics?
    • Chris Auld’s Remarks on Chen and Pearl (2013) Regression and Causation: A Critical Examination of Six Econometrics Textbooks.
    • Friends don’t let friends do IV…but…Friends do let friends do IV
    • If correlation doesn’t imply causation, then what does?

Disciplina: ECONOMETRIA AVANCADA 2015

Período Letivo: 2015/1

Professor: Hedibert Freitas Lopes – www.hedibert.org

Monitor: Paloma Vaissman Uribe

Objetivo: O objetivo do curso é apresentar os conceitos e métodos de análise multivariada de dados, aplicando-os a dados reais e interpretando os resultados de forma prática. No curso de análise multivariada são utilizados conceitos de estatística básica e inferência, com ênfase na resolução de problemas reais e interpretação dos resultados.  Na maioria dos estudos, a complexidade dos fenômenos estudados faz com seja necessário coletar informações sobre um conjunto de variáveis. A análise multivariada permite o estudo simultâneo de um conjunto de variáveis, aproveitando a estrutura de correlação existente entre as mesmas.  Nesta disciplina são apresentadas técnicas de análise de dados quantitativos e qualitativos, discutindo aplicações nas áreas de marketing, operações, recursos humanos e finanças.

Programa de Ensino 

Introducao ao R by Paloma Uribe (Material apresentado na 1a monitoria) 

Lista de exercicios

• • • 19/02/2015: Primeira lista (Solucao)
    • 05/03/2015: Segunda lista (Codigo R) (Solucao)
    • 24/04/2015: Terceira lista (Codigo R) (Solucao)
    • 05/05/2015: Quarta lista  (Codigo R) (Solucao)
    • 28/05/2015: Quinta lista (exercicio resolvido)

Trabalhos em grupo

• • • 05/03/2015 (entrega 19/03/2015): Trabalho 1
    • 19/03/2015 (entrega 02/04/2015): Trabalho 2
    • 30/04/2015 (entrega 23/04/2015): Trabalho 3
    • 02/06/2015 (entrega 23/04/2015): Trabalho 4

Prova intermediaria: Prova (solucao)

Prova final: Solucao

Notas de aula

• • • PARTE I: Apresentacao + Introducao
    • PARTE II: Modelos AR, MA & ARMA
    • PARTE III: Metodologia Box & Jenkins
    • PARTE IV: Modelos lineares nao-estacionarios
    • PARTE V: Previsao + Previsao de AR(1) e AR(1)
    • PARTE VI: SARIMA + SARIMA(1,1,1)x(1,1,1)
    • PARTE VII: Modelos lineares dinamicos
    • PARTE VIII: Modelos ARCH e GARCH
    • PARTE IX: Tendencia-estacionario ou diferenca-estacionario?
    • PARTE X: Modelos SV-AR(1)
    • PARTE XI: Modelos Autoregressivos Vetoriais (VAR)
    • PARTE XII: Cointegracao, regressao espuria, VAR+ECM

Codigo R

• • • PARTE I: airline – indicepluviometrico – loadfactor – macro – petrobras – exercicios
    • PARTE IV: Random walk with drift – Teste ADF – Nelson&Plosser – Nelson&Plosser paper
    • PARTE V: forecasting-airline – forecasting-GDP
    • PARTE VI: Local level model (stey-by-step) + dynamiclinearmodels.R
    • PARTE VIII:  Trend stationarity or not?
    • PARTE IX: SV-AR(1) in R + Volatility Index (VIX) versus SV-AR(1)
    • PARTE X: I Exemplo de VAR + II Exemplo de VAR

Textos complementares

• • • PARTE VI:
      • Petris & Petroni (2011) State space models in R
      • Tusell (2011) Kalman filtering in R
      • Journal of Statistical Software (2011) – Volume 41 – Statistical Software for State Space Methods
    • PARTE VII: Glossary to ARCH (GARCH) by Tim Bollerslev
    • PARTE VIII:  A critique of the application of unit root tests (John Cochrane, 1991)

Conjuntos de dados

• • • airline.txt
    • presidenteprudente.txt – ribeiraopreto.txt
    •  american.txt – delta.txt – united.txt 
    • macro.csv
    • petrobras.csv
    • nelsonplosser-data.txt
    • UKdriversKSI.csv
    • vix-sp500.csv
    • sp500.csv

Alguns sites interessantes para o curso

• • • Pacote estatistico R
    • Pagina do livro de Morettin & Toloi
    • Pagina do livro de Shumway & Stoffer
    • Pagina do livro de Zivot & Wang
    • Pagina do livro de Tsay

Disciplina: ECONOMETRIA 2014 – Turma 4ECO (Economia)
Período Letivo: 2014/1

Professor: Hedibert Freitas Lopes – www.hedibert.org

Objetivo: Apresentar uma abordagem introdutória a Econometria dando ênfase tanto à base estatística quanto a aplicações econômicas.  Será discutido, em detalhes, o significado e as implicações das suposições do modelo linear geral. Ainda, serão descritos e aplicados testes de violações das hipóteses do modelo linear geral, bem como serão apresentados e aplicados estimadores alternativos ao de mínimos quadrados ordinários (MQO). Ao final desse curso, o aluno deverá ser capaz de utilizar técnicas estatísticas adequadas para mensurar quantidades de interesse e realizar previsões.

Programa de Ensino 

12/02/2014: Apresentacao do curso + Primeiro exemplo

14/02/2014: Regressao Linear Simples – Parte 1: Minimos Quadrados Ordinarios & R2

• • • VendedoresRH.xls
    • Mais sobre minimos quadrados ordinarios
    • Modelando salario via anos de experiencia, posicao e sexo (Codigo R)  (Dataset)

19/02/2014: Regressao Linear Simples – Parte 2: formas funcionais  

21/02/2014: Regressao Linear Simples – Parte 3: Suposicoes, propriedades e teste-t

26/02/2014: Regressao Linear Multipla – Parte 1: Estimacao

• • • Revisao de Algebra Linear Basica para Regressao Linear (Desenvolvido pela Professora Marie Davidian, Department of Statistics, NCSU)

28/02/2014: Regressao Linear Multipla – Parte 2: R2 ajustado

• • • Bank wages: Mais sobre R2 e R2 ajustado para 15 modelos (Codigo R)  (Dataset)

07/03/2014: Regressao Linear Multipla – Parte 3: Suposicoes e propriedades

• • • Data on monthly earnings, education, several demographic variables, and IQ scores for 935 men in 1980. (Dataset)  (Codigo R)
    • Data on 4,137 US college students.(Dataset)  (Codigo R)

12/03/2014: Regressao Linear Multipla – Parte 4: Inferencia

• • • Tutorial em R – Parte I (Datasets: salario.txt & salario.csv)

14/03/2014: Regressao Linear Multipla – Parte 5: Interacao e funcao quadratica

• • • Tutorial em R – Parte II :(Dataset: houseprices.txt)

19,21&26/03/2014: Regressao Linear Multipla – Parte 6: Informacao qualitativa atraves de variaveis dummy

• • • Precos de imoveis via dummies – houseprices.txt
    • Precos de imoveis via dummies & interacoes – houseprices.txt  – Codigo R
    • Salarios de um banco americano – bankwages.txt – Codigo R
    • Precos de sucos de laranja – orangejuice-chicagoarea.csv – Codigo R
    • Precos de Toyota Corollas usados – toyotacorolla.csv – Codigo R
    • Peso vs altura, idade, sexo – peso-altura-idade-sexo-criancas.txt – Codigo R
    • Monthly earnings, education, IQ, etc for 935 men in 1980 – wage2-wooldridge.txt – Codigo R

11/04/2014: Regressao Linear Multipla – Parte 7: Teste F parcial

• • • Lista de exercicios – resolva a prova intermediaria
    • Prova intermediaria – Analise exploratoria
    • Conjuntos de dados: temco.txt – temcoprod.txt – gpa2.txt
    • Codigo R para o calculo do Teste F parcial (Exemplo das notas de aula)

16/04/2014: Regressao Linear Multipla – Parte 8: Teoria assintotica

• • • Simulando regressoes lineares simples – Codigo R

23/04/2014: Regressao Linear Multipla – Parte 9: Teste do multiplicador de Lagrange

• • • R code for hprice1 example
    • Sawyer (2002) The Method of Lagrange Multipliers.
    • Buse (1982) The Likelihood Ratio, Wald, and Lagrange Multiplier Tests: An Expository Note.  The American Statistician, Vol. 36, No. 3, Part 1 (Aug., 1982), pp. 153-157.
    • Engle (1983) Wald, Likelihood Ratio, and Lagrange Multiplier Tests in Econometrics.  In Intriligator and Griliches. Handbook of Econometrics II. Elsevier. pp. 796–801.

25/04/2014: Regressao Linear Multipla – Parte 10: Regression specification error test  (RESET)

• • • Codigo R para o RESET em dados simulados
    • RESET para modelos nivel-nivel e log-log de precos de residencias – hprice1.txt (Codigo R)
    • RESET: dados de 5848 municipios brasileiros (Codigo R)

30/04/2014: Real-time analysis of three datasets

• • • Indicadores socio-economicos municipais brasileiros (Codigo R)(Arquivo XLS)
    • Performance media dos alunos de diferentes escolas (Codigo R)
    • Data on 177 chief executive officers (Codigo R)

07-09/05/2014: Regressao Linear Multipla – Parte 11: Homocedasticidade

• • • Airline data (Codigo R)
    • Car mileage (Codigo R)

14/05/2014: Trabalho em grupo

• • • Analfabetismo vs renda per capita nos municipios brasileiros (Codigo R)
    • House price in Windsor, Ontario, Canada (Codigo R)
    • Taxa anual de homicidio – jovens de 15 e 29 anos (in 100000 habitantes entre 1980 e 2009) nos 12 municipios brasileiros sedes da copa do mundo de 2014 (Codigo R)

16 a 30/05/2014:  Endogeneidade & Equacoes simultaneas

• • • Return to education: simple IV regressions (women) (men) (Angrist and Krueger)
    • ARP-Analise-e-Resolucao-de-Problemas  (Dataset)
    • Simulating IV scenarios  (R code)

Listas de exercicios + gabaritos

• • • Lista 1  (solucao)
    • Lista 2  (solucao)
    • Lista 3  (solucao)
    • Lista 4  (solucao)
    • Lista 5  (solucao)
    • Lista 6  (solucao)

Conjuntos de dados

• • • VendedoresRH.xls
    • salario.txt: Salario vs posicao, anos de experiencia e sexo
    • bankwages.txt: Salario vs salario inicial, educacao, sexo, minoria e categoria
    • wage2-wooldridge.txt: Monthly earnings, education, IQ, etc for 935 men in 1980
    • gpa2-wooldridge.txt: Data on 4,137 US college students
    • houseprices.txt: House prices vs offers, sqft, brick, bedrooms, bathrooms and neighborhood
    • orangejuice-chicagoarea.csv: Weekly sales of 64oz orange juice containers in the Chicago area
    • toyotacorolla.csv: Sales prices and vehicle characteristics of 1436 used Toyota Corollas
    • peso-altura-idade-sexo-criancas.txt: peso vs altura, idade e sexo
    • gpa2.txt: same as gpa2-wooldridge.txt
    • temco.txt
    • temcoprod.txt
    • hprice1.txt: dados sobre 88 residencias.
    • dadosmunicipais.csv: dados de 5848 municipios brasileiros (analfabetismo, renda, desigualdade).
    • caliescom.xls: Performance media de 420 escolas em um teste padronizado.
    • ceosal2.xls: Data on 177 chief executive officers.
    • florida2000.xls: Florida 2000 presidential vote as of 5pm ET Saturday, Nov. 11, 2000 (after recount).
    • temcoprod.xls: Caracteristicas de funcionarios de uma firma.
    • simulados_producao.xls: Modelagem da produção de firmas de um determinado setor.

Course: Business Statistics 41000-81/82 – Spring Quarter 2013

Professor: Hedibert Freitas Lopes

Teaching Assistant: Samir Warty