A three-part review of methods for selecting controls for case-control studies was published, and methods for analyzing a """"""""partial questionnaire design"""""""" to reduce the burden of questions in case-control studies, have been developed. Progress has been made in developing logistic regression analyses for case-control studies with cluster sampling of controls and for analyzing group randomized intervention trials with adjustment for covariate imbalances and missing data. Flexible methods of analysis of case-control studies that treat unsampled subjects as """"""""missing data"""""""" have also been developed. A study of the efficiency of adjustments on """"""""propensity"""""""" to control for confounding showed that this method was less efficient than standard methods. Pseudolikelihood methods for correcting case-control analyses with errors in covariates were published, but separate work showed that such adjustments based validation data can be misleading if the """"""""gold standard"""""""" is subject to error. Methods for estimating attributable risk in the presence of error-prone exposure measurements were described. Methods for inference on estimates of absolute risk were developed, and a computer program was written to project breast cancer risk with appropriate confidence intervals on risk estimates. Validation studies for this model are in progress. Methods for analyzing repeated categorical data were developed that allow for a proper treatment of missing data, even when missingness is not completely at random. A paper giving a method for estimating treatment effects in clinical trials in the presence of non-compliance was described, and a related application to observational studies is in press. Spline models have been developed for analyzing the incubation distribution hazard and HIV incidence curves in AIDS research. Backcalculation methods have been generalized to allow for treatment effects, and, in separate work, to permit age-specific estimates of HIV infection curves. A paper is in press describing non-parametric methods for estimating the rate of HIV transmission from mother to infant, and biases were noted in methods of analysis used to infer effects of treatment from observational data.
