B0554
Title: Anytime-valid linear models and regression adjusted causal inference in randomized experiments
Authors: Michael Lindon - Netflix (United States) [presenting]
Abstract: Linear models are commonly used in causal inference for the analysis of experimental data. There is, however, a replicability crisis in applied research through unknown reporting of the data collection process. In modern A/B tests, there is a demand to perform regression-adjusted inference on experimental data in real time. Together, these motivate modernizing linear model theory by providing "Anytime-Valid" inference. These replace classical fixed-n Type I error and coverage guarantees with time-uniform guarantees, safeguarding applied researchers from p-hacking, allowing experiments to be continuously monitored and stopped using data-dependent rules. With an emphasis on experimental data, it can relax the linear model assumption in randomized designs. In particular, completely nonparametric confidence sequences are provided for the average treatment effect in randomized experiments, without assuming linearity, Gaussianity or no omitted variables. A particular feature of contributions is their simplicity. The test statistics and confidence sequences have closed-form expressions of the original classical statistics, meaning they are no harder to use in practice. This means that published results can be revisited and reevaluated, and software libraries which implement linear regression can be easily wrapped.
