Statistics Colloquium

Title:  Predictors of The Variance of a Finite population Under a location and scale parameters super-population model

Abstract: The lecture starts with comments about the two main approaches to sampling from finite populations. This lecture focuses on the modelling approach, in which we predict the variance of a finite population. We start with discussing the structure of predictors, which are based on  given sample values and a super population model of the population variables. We then show the structure of equivariant predictors, the maximal invariant statistics, and the properties of the risk function under invariant loss function. Finally we show how to obtain the minimum risk predictor and provide two examples.