NATIONAL BUREAU OF ECONOMIC RESEARCH, INC.

SI 2013 Econometrics Lectures: Econometric Methods for High-Dimensional Data

Victor Chernozhukov, Matthew Gentzkow, Christian Hansen, Jesse Shapiro, Matthew Taddy

July 15-16, 2013

Hotel Marlowe

25 Edwin H. Land Boulevard

Cambridge, Massachusetts

PROGRAM

Monday, July 15:
8:15 am 8:45 am	Coffee and pastries Matthew Taddy, University of Chicago Prediction with high-dimensional data (1) (slides) Metrics: False discovery, predictive performance, model probabilities Model choice: FDR control, cross-validation, information criteria Algorithms for finding candidate models, forward stepwise regression
10:15 am	Break
10:45 am	Matthew Taddy, University of Chicago Prediction with high-dimensional data (2) Penalized maximum likelihood and sparse regularization Lasso model fit and penalty selection Properties and performance of regularization methods
12:15 pm	Lunch
1:30 pm	Matthew Taddy, University of Chicago Introduction to factor models Principal components analysis and regression Selection for Factor Models
2:30 pm	Break
2:45 pm	Matthew Gentzkow, University of Chicago and NBER Jesse Shapiro, University of Chicago and NBER Applications: Using text as data (slides) * Partisanship in the news media * Sentiment in financial news * Topic modeling
4:45 pm	Adjourn
Tuesday, July 16:
8:15 am 8:45 am	Coffee and pastries Victor Chernozhukov, Massachusetts Institute of Technology (slides) Estimating treatment effects with high-dimensional data (1): Selecting instruments
10:15 am	Break
10:30 am	Victor Chernozhukov, Massachusetts Institute of Technology Estimating treatment effects with high-dimensional data (2): Selecting control variables
12:00 pm	Lunch
1:30 pm	Christian Hansen, University of Chicago Applications: Estimating treatment effects with high-dimensional data (slides) * Selecting control variables * Modeling trends in panel differences-in-differences * Selecting from a large set of randomized instruments
3:30 pm	Break
3:45 pm	Matthew Gentzkow, University of Chicago and NBER Jesse Shapiro, University of Chicago and NBER Nuts and bolts: Computing with large data (slides) * Collaborative computing * Databases * Distributed computing
5:15 pm	Adjourn

REFERENCES

Prediction:

Hastie, Trevor, Robert Tibshirani and Jerome Friedman. 2009. Elements of Statistical Learning: Data Mining, Inference and Prediction. Second Edition. http://www-stat.stanford.edu/~tibs/ElemStatLearn/

Text Mining:

Taddy, Matthew. 2013. “Multinomial Inverse Regression for Text Analysis.” Journal of the American Statistical Association, forthcoming. http://arxiv.org/abs/1012.2098

Blei, David M., and John D. Lafferty. 2007. “A Correlated Topic Model of Science.” The Annals of Applied Statistics 1 (1) (June 1): 17–35. http://www.jstor.org/stable/4537420

Treatment Effects:

Alexandre Belloni, Victor Chernozhukov, Christian Hansen, "Inference for High-Dimensional Sparse Econometric Models", Advances in Economics and Econometrics, 10th World Congress of Econometric Society, 2010 (http://arxiv.org/pdf/1201.0220v1.pdf) 

Belloni, Alexandre, Daniel Chen, Victor Chernozhukov and Christian Hansen. 2012. “Sparse Models and Methods for Optimal Instruments with an Application to Eminent Domain.” Econometrica 80(6): 2369-2430. http://onlinelibrary.wiley.com/doi/10.3982/ECTA9626/pdf

Belloni, Alexandre, Victor Chernozhukov and Christian Hansen.  2013.  “Inference on treatment effects after selection amongst high-dimensional controls.” CEMMAP, http://www.cemmap.ac.uk/wps/cwp261313.pdf

Leeb, Hannes and Benedikt M. Pötscher. 2008. “Can one estimate the unconditional distribution of post-model-selection estimators?” Econometric Theory 24(2): 338–376. http://dx.doi.org/10.1017/S0266466608080158

Code and Data

Gentzkow, Matthew and Jesse Shapiro. 2013. Code and Data for the Social Sciences: A Practitioner’s Guide. http://faculty.chicagobooth.edu/jesse.shapiro/research/CodeAndData.pdf

LINKS TO CODE FOR EXAMPLES USED IN LECTURES

Prediction

semiconductor examples: http://faculty.chicagobooth.edu/matt.taddy/teaching/scripts/semiconductors.R
class on factor analysis: http://faculty.chicagobooth.edu/matt.taddy/teaching/scripts/07Factors.R
In general, all data and code is on my teaching page:  http://faculty.chicagobooth.edu/matt.taddy/teaching
And the penalized regression software is here: http://www.cran.r-project.org/web/packages/gamlr

Treatment Effects

http://www.mit.edu/~vchern/#veryhigh
http://faculty.chicagobooth.edu/christian.hansen/research/#Code