In causal inference, the goal is often to estimate average treatment effects. Selecting a model by cross-validation in this context can be problematic, as models that exhibit great predictive accuracy can be suboptimal for estimating the parameter of interest. We discuss several approaches to perform model selection in this context and compare their performance on simulated data sets.
- Speaker: Dominik Rothenhaeusler (Stanford University)
- Friday 30 October 2020, 16:00–17:00
- Venue: https://maths-cam-ac-uk.zoom.us/j/92821218455?pwd=aHFOZWw5bzVReUNYR2d5OWc1Tk15Zz09.
- Series: Statistics; organiser: Dr Sergio Bacallado.