Reproduce “Sequential Kernel Embedding for Mediated and Time-Varying Dose Response Curves” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Rahul Singh, Liyuan Xu, Arthur Gretton
The authors develop a sequential kernel embedding with kernel ridge regression to estimate continuous-treatment effects across time while retaining treatment-confounder feedback. It represents covariate distributions and estimates dose sequences. In Job Corps, fourth-year employment is about 64% for a high-response schedule versus 56% with no class hours, conditional on identifying assumptions; at 5,000, simulations favor it over alternatives.
Key insights
- Sequential kernel embedding provides a method to transform complex sequential integrals into tractable inner product evaluations within RKHS, bypassing the need for explicit density estimation of high-dimensional conditional distributions.
- The RKHS framework's property of bounded evaluation functionals enables robust nonparametric estimation and theoretical analysis for continuous treatments in sequential causal settings.
- The framework allows for the preservation of nonlinearity, dependence, and effect modification, offering a more general approach than methods relying on restrictive assumptions.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.