Weighted Logrank Tests for Interval Censored Data When Assessment Times Are Related to Treatment but Not to Event Time

• Activity Number: 506
• Type: Contributed
• Date/Time: Wednesday, August 3, 2011 : 10:30 AM to 12:20 PM
• Sponsor: Biometrics Section
• Author(s): Michael P. Fay*+ and Joanna H. Shih
• Companies: National Institute of Allergy and Infectious Diseases and National Cancer Institute
• Keywords: interval censoring ; logrank test ; permutation test ; small sample size ; imputation

Abstract

We consider weighted logrank (WLR) tests for interval censored data when assessment times may depend on treatment and for each individual we only observe two assessment times, the one before and the one after the event of interest. We review that for large samples with a small number of unique assessment times, the score test under the grouped continuous model (GCM) retains the type I error even under assessment-treatment dependence. Although the permutation form of the WLR test under the GCM is theoretically valid under assessment-treatment independence, we give heuristics and simulations to show that even under assessment-treatment dependence this type of WLR test approximately retains the type I error in all but some very extreme cases. Next, we study and extend the imputation WLR tests of Huan, Lee, Yu (2008, Stat in Med), showing that the distribution of the rank-like scores asymptotically do not depend on the assessment times. We confirm through simulation that the imputation WLR tests retain the type I error even when the number of individuals in each treatment group are small and possibly unbalanced in size and with assessment-treatment dependence.