Research in Likelihood-based Diagnostics and Influence
In regression diagnostics interest has focussed on developing the links between deletion and local influence. Within deletion influence a conditional approach to the problem of masking was developed, and local influence diagnostics for Box-Cox transformation were derived, a subject of earlier inferential work. Current interests in both deletion and local influence focus on a unified likelihood approach to influence on goodness of fit tests and on model testing, rather than for parameter estimation as is more conventional. Particular attention is being given to binomial regression.
References
Regression transformation diagnostics using local influence. J. Amer. Stat. Assoc. 83, (1988), 1067-1072.
Local and deletion influence. In Directions in Robust Statistics and Diagnostics, Part 1, (1991) W. Stahel and S. Weisberg (eds.), 141-157, Springer-Verlag.
Deletion influence and masking in regression. J.R.Statist.Soc.B, 57, (1995); 181- 189.
Theory and illustration of regression influence diagnostics; with G.C.Brown, Communications in Statistics: Theory and Methods (2000), 29 (9), 131-160.