Jan Scherer - Uni Bonn

"Inference in high-dimensional partially linear sparse models"


Abstract

We construct a testing procedure for average treatment effects with many controls. This problem has been studied by many authors under the assumption that the data follows a high-dimensional sparse linear model. These models restrict the functional form of the treatment effect to be linear. We weaken this assumption and model the data as a high-dimensional sparse partially linear model which allows for non-linear treatment effects. Our proposed testing procedure allows to infere the locations and signs of non-zero treatment effects.


Additional information:

  • Speaker: Jan Scherer
  • Time: Thursday, 16.12.2021, 16:00 - 17:00
  • Location: Online via Zoom
  • Further links:
  • Organizer: Statistics Group
  • Contact:

Wird geladen