Mukhopadhyay (1977, 1986) considered conditions under which the model - dependent optimal predictor T *(0,l;xg) becomes bias - robust in a class of predictors under alternative polynomial regression models ξ(δ0. .... δj; v(x )). He obtained a post- sample bias - robust estimator under this class of alternatives. In this note we examine his results, further, for choosing a bias - robust estimator T *(0,l;xh) for the class of alternative models ξ(1,1; v (x)) by considering empirical results from simulated populations.