B1257
Title: Asymptotic inference for the mean with minimal assumptions
Authors: Siddhaarth Sarkar - Carnegie Mellon University (United States) [presenting]
Arun Kuchibhotla - Carnegie Mellon University (United States)
Abstract: The central limit theorem provides asymptotic confidence intervals for the mean of a distribution from an IID sample, only requiring second-moment conditions, among other possible conditions. However, the theorem doesn't hold under more generalized settings, such as heavy tails. A method is proposed to construct an asymptotically valid confidence interval for the mean, assuming weaker conditions on the distribution. The approach involves using the confidence set for the CDF of the distribution and inverting a distribution-free statistic. The finite sample properties of the proposed confidence interval are examined, and its performance is compared to other methods in the literature.
