Relatore:
Simoni Anna
- CREST and CNRS
mercoledì 15 febbraio 2017
alle ore
12.30
Polo Santa Marta, Via Cantarane 24, Room 1.59
We consider inference for a linear regression model where the number of covariates is very large, possibly much larger than the sample size n, and where some covariates are allowed to be endogenous. The model is identified and estimated using instrumental variables. We are interested in developing a valid inference procedure that allows to construct confidence intervals and hypothesis tests for the low-dimensional subparameter of interest. We propose a three-step estimation procedure based on a LASSO type estimator, suitably modified to have a tractable limiting distribution. The estimator that we obtain is no longer sparse but is consistent and approximately Gaussian. We derive confidence intervals and hypothesis testing procedures. (Joint with Christoph Breunig and Enno Mammen)