Presented by: 
Siuli Mukhopadhyay - Indian Institute of Technology, Bombay
Tue 27 May, 11:00 am - 12:00 pm

Abstract: In this talk the problem of model misspecification in generalized linear models (GLMs), particularly multinomial models, and its effect on design selection is studied.  In case of GLMs, model misspecification may occur when the form of the linear predictor and/or the link function is incorrect. To address the issue of a misspecified link a parametric family of link functions for the multinomial response model is proposed. The link family includes the multicategorical logistic link as one of its members. Conditions for the local orthogonality of the link and the regression parameters are given. It is shown that local orthogonality of the parameters in a neighbourhood makes the link family location and scale invariant. Confidence regions for jointly estimating the percentiles based on the parametric family of link functions are also determined.