Problems in High-dimensional Econometrics

Mitarbeiter in diesem Projekt:

In the recent years, large data sets became available for researchers. In order to analyse such data sets in which the number of regressors is very large compared to the number of observations (or even larger), new techniques are in need. Whithin the research project, boosting, a technique originally developed in Machine Learning and now introduced in Statistics, should be applied to Econometric problems. Moreover, applications of Lasso are also analyzed. In a first paper (Mittnik, Robizonov and Spindler, 2013), boosting is applied to volatility modelling. A further application is IV estimation with many instruments. Additionally, the problem of significance should be addressed which has been neglected in the past. A research stay related to the project was funded by the DFG and conducted at MIT, Cambridge, USA, on invitation of Prof. Chernozhukov. During this stay two joint projects were started and are still in progress. A first publication resulting from this project is "Lasso for Instrumental Variable Selection", forthcoming in the Journal of Applied Econometrics.

Aus diesem Projekt hervorgegangene Publikationen:
  • Spindler, Martin; Luo, Ye (2017): L2-Boosting for Economic Applications, American Economic Review Papers & Proceedings, 107, 5, 270-273
  • Spindler, Martin; Chernozhukov, Martin; Hansen, Chris (2016): hdm: High-dimensional Econometrics, R Journal, 8/2, 185-199

News & Events

SHARE User Conference 2019 in Budapest, Hungary

Registration open


SHARE Buchvorstellung in Brüssel - 25. Juni 2019

Neueste Studienergebnisse des Survey of Health, Ageing and Retirement



Latest Newsletter

MEA is coordinating