This application proposes extension of the investigator's previous work on the analysis of longitudinal studies arising in cancer research. The approach is concerned with both randomized clinical trials and epidemiological investigations. In particular, the applicant proposes to address the issue of adjusting for confounders that are observed post-randomization in randomized clinical trials. The applicant has previously postulated the view that since the confounder is observed post-randomization, it should be treated as an outcome variable and analyzed accordingly, rather than regarded as a covariate or predictor value. This approach will be extended in the proposed work with the following specific goals: 1) Developing methods for adjusting for survival confounding measures in repeated measures analysis. 2) Developing methods to deal with informative censoring in non-ignorable non-response in longitudinal trials. 3) Conduct a simulation study of the conditional inference procedures of the methodology. 4)Develop procedures for analyzing bivariate repeated measures data. 5)Develop procedures for performing sample size calculations for repeated measures with both discrete and continuous outcomes.
