Survival Analysis Without Survival Data: Estimation by Comparing Explanatory Variables from Unbiased and Length-Biased Samples

• Activity Number: 286
• Type: Topic Contributed
• Date/Time: Tuesday, August 3, 2010 : 8:30 AM to 10:20 AM
• Sponsor: Biometrics Section
• Author(s): Kwun Chuen Gary Chan*+
• Companies: University of Washington
• Keywords: accelerated failure time model ; case control ; empirical likelihood ; exponential tilted model ; prevalent cohort ; mean residual time

Abstract

It is typical in regression analyses that marginal distribution of covariates does not contain any information about regression parameters of interest when sampling mechanism is unbiased. Under length biased sampling, however, covariate values associated with larger values of outcome are preferentially sampled. We show that by comparing covariate distributions from an unbiased and a length biased samples, estimating equations for slope parameters in regression models can be constructed without using information from the outcome of interest. An implication to survival analysis is that inference of relative risk parameters are possible based on covariate data collected from an incident cohort and a cross-sectional sampled prevalent cohort, even when there is no prospective follow-up to collect survival data.