The Effects of Model Misspecification, Conditions for Robustness, and Bayesian Modeling

Abstract

As seen in the previous chapters, the optimality of predictors of population quantities might strongly depend on the particular model under which they have been derived. For example, if one assumes model SM2 (model ψ), as the working model, the ψ-BLUP of the population total is the ratio predictor

$${\hat T_R} = n{\bar y_s} + \frac{{{{\bar y}_s}}}{{{{\bar x}_s}}}\left( {N - n} \right){\bar x_r}$$