The Mathematical Sciences Department invites you to a Statistics colloquium given by Dr. Tathagata Bandyopadhyay of Indian Institute of Management, Ahmedabad, India (currently visiting Michigan State)
Estimation of finite population distribution function (hereafter FPDF) is considered to be an important problem in survey sampling since it summarizes almost all the relevant information about the finite population characteristics of interest. Chambers and Dunstan (1986) (hereafter CD) develop a predictive estimator of FPDF using linear regression model in the super-population to incorporate unit level information in the population on a set of auxiliary variables. Here we extend CD’s estimator using a nonparametric regression model in the super-population based on recently developed penalized splines (P-splines) regression and find it’s asymptotic model bias and variance. The proposed estimator performs better than CD’s predictive estimator if the linearity assumption of the super-population model fails.
* This is a joint work with Sumanta Adhya, Gourangadeb Chattopadhyay
Coffee will be served in FO 2.610F at 1:30 PM.