Christopher

Cox

,

PhD

I am interested in the efficient computation of maximum likelihood estimates for generalized nonlinear models, and in the practical application of these models. An important aspect of this research is the application of the delta method to provide standard errors for nonlinear functions of model parameters. A recent paper employed the delta method to provide a standard error for a new definition of the limit of quantitation for a laboratory assay. A second paper involved a simplified computation of the standard error for a model-based estimate of the total and partial attributable risk (AR) under both case-control and cross-sectional sampling. Current research involves providing explicit formulas for the variance of the AR based on the delta method. An important aspect of this work is the use of a Poisson model to account for the estimation of the distribution of exposure and covariates. I have collaborated with Dr. Muñoz on a study of the generalized gamma (GG) family of parametric survival models, and its application in a period analysis to study the effect of HAART therapy on the time from AIDS to death. Graphical tools for the presentation of relative times and relative hazards are an important aspect of this work. Variability of these estimates was computed using the delta method. A second paper determined the hazard behavior of the generalized F distribution and proposed this larger family as a check on the fit of the GG. A third generalized the AR to the survival context and proposed the concepts of attributable survival and attributable survival time.

I am a member of the MACS Behavior Working Group and am involved in several collaborative projects in both MACS and WIHS, as well as CKID. In addition I am working with Dr. João Lima on clinical imaging studies in the School of Medicine.