The day in conversation for 2026-11-08
discussion
Transcript
The discussion: Iris: The primary takeaway from the causal inference work is that these debiased tests move beyond relying on strict ignorability assumptions when dealing with high-dimensional outcome spaces, specifically how they handle joint distributional effects. Ruth: I agree that moving away from joint Wald tests when signal strength is low is key, but I’m curious how robust these procedures remain if the underlying unconfoundedness assumption itself is mildly violated by measurement error in the covariates? Warren: If we look at those methods—conditional randomization, double-residualization, and AIPW—they seem designed to isolate the structural causal effect even when distributional assumptions are shaky, but I wonder if their computational efficiency gain actually outweighs the complexity introduced by needing to estimate those residualized components reliably. Iris: That relates directly to what we saw last week with the high-dimensional econometrics package; the trade-off is usually between parameter estimation variance and asymptotic
More episodes
- Daily Summary for 2026-11-19
- 2608.21274-Recommendation Quality and the Concentration of Consumption: Experimental Evidence from Netflix
- 2408.02757-A nonparametric test for diurnal variation in spot correlation processes
- 2306.16393-High-Dimensional Canonical Correlation Analysis
- 2605.17299-Geometric Brownian motion with intermittent entries and exits
- 2607.04753-Fooling Yourself: how narratives shape beliefs
- 2408.14671-Inference on Treatment Effects with Repeatedly Mismeasured High-Dimensional Controls
- 2201.06898-Difference-in-Differences Estimators for Treatments Continuously Distributed at Every Period
- 2501.15307-Influence Functions: A Unified Differentiation-Based Framework
- The day in conversation for 2026-11-19