3 p.m. - 4 p.m. Location: FO 2.702
Sy Han (Steven) Chiou
Department of Biostatistics
Harvard T.H. Chan School of Public Health
Permutation tests for general dependent truncation
Quasi-independence is a common assumption for analyzing truncated survival data that are frequently encountered in biomedical science, astronomy, and social science. While the concept of censoring has been rigorously studied, many are not aware of the analytic issues that arise with delayed entry, or general truncation. Ignoring dependent truncation can lead to severely biased estimation and inference. Current methods for testing quasi-independent truncation are powerful for monotone alternatives, but not otherwise. We extend methods for detecting highly non-monotone and even non-functional dependencies and develop nonparametric tests for dependent truncation that are powerful against non-monotone alternatives. We compare computation time, size and power of both conditional and unconditional permutation procedures. We apply our results to a study on the cognitive and functional decline that had delayed entry due to post-baseline imaging.
Sponsored by the Department of Mathematical Sciences