Mathematical Sciences: Computing Model Robust Statistics

This research aims to extend model robust statistical procedures to handle data with general covariance structure, for example, time series and spatial data. The focus of the research on non independent identically distributed random variables is computationally intensive and offers tools to merge previous approaches. Methods of crossvalidation and robust maximum likelihood will be generalized. This research is in the general area of statistical analysis of dependent observations and will impact procedures for understanding such data including radiocarbon dating, nuclear accountability tank calibration, and software development effort prediction.