PhD in Deletion and Local Influence for Fit
This topic is in the diagnostics area of statistical modelling and inference. There may be concern that a small part of a data set, or a supporting model assumption, is crucial to an aspect of the inference. The particular aspect will depend on circumstances, and could be the parameter estimates. Perhaps the deletion of a particular case of the data causes the parameter estimates to change appreciably. This situation has been explored fairly thoroughly, but whether the fit changes appreciably at the same time is another question, and sometimes of more concern. However, the deletion of data is sometimes considered too drastic an effect, and the more mathematical and geometric local influence was developed to allow data or assumptions to be slightly perturbed, and the influence of these ascertained. This concern will form an important part of the research area. The development to one or two particular methodological areas, such as repeated measures analyses and random effect models is also likely to be a component of the work.