Talk
Speaker: Prof. Monika Bhattacharjee, IIT Bombay
Title: Testing High-Dimensional Means under Sparse Missingness
Time, day and date: 4:00:00 PM – 5:00:00 PM, Monday, March 30
Venue: Room 215 (https://meet.google.com/dkp-hbsd-rdd)
Abstract: We propose a new testing procedure for the population mean vector in high-dimensional settings with missing observations. Existing approaches typically require finite fourth moments and impose restrictive assumptions on the missingness mechanism, often effectively limiting attention to dense missing patterns.
Our method overcomes these limitations by standardizing the test statistic using the conditional variance given the observed missingness structure. This removes the need for the missingness probabilities to be bounded away from zero and allows for genuinely sparse missing patterns. Moreover, the proposed procedure requires only slightly stronger than second moments, making it applicable to heavy-tailed distributions where the fourth moment may not exist.
The framework further accommodates weak dependence across variables through geometric alpha-mixing and does not rely on the assumption that the observations are identically distributed, thereby substantially broadening its applicability. We establish the asymptotic normality of the test statistic, along with consistency and correct asymptotic size under suitable conditions.