Locally Efficient Estimation for Current Status Data

If one only observes whether a failure time variable, T, exceeds a monitoring time, C, then the structure is called current status data. We consider this data structure extended to allow the presence of covariates, Z, and time-dependent covariate processes observed until C. We present a one-step estimator of the parameters of the usual linear regression model of T on Z that allows for dependent monitoring. Under conditions our estimator is consistent, asymptotically normal, and efficient. The ideas are presented via the analysis of data from an experiment designed to investigate the relationship between the time until development of a tumor and the dose level of a suspected tumorigen. Study rats were randomly allocated to dose groups and weighed repeatedly until sacrifice.