Distribution-Free Conditional Median Inference
share
We consider the problem of constructing confidence intervals for the median of a response Y ∈ℝ conditional on features X = x ∈ℝ^d in a situation where we are not willing to make any assumption whatsoever on the underlying distribution of the data (X,Y). We propose a method based upon ideas from conformal prediction and establish a theoretical guarantee of coverage while also going over particular distributions where its performance is sharp. Further, we provide a lower bound on the length of any possible conditional median confidence interval. This lower bound is independent of sample size and holds for all distributions with no point masses.
READ FULL TEXT
