Generalized Minimum Bias Models

In this research, we propose a flexible and comprehensive approach for minimum bias models -- "Generalized Minimum Bias Models"(GMBM). Unlike the Generalized Linear Models (GLMs) that require the exponential family distribution assumption of response variables, the GMBM approach relaxes the distribution assumption. In addition, due to its model selection flexibility, we believe that GMBM will improve the accuracy and the goodness of fit of classification rates. All the multiplicative minimum bias models published to date and the commonly used multiplicative GLMs (such as Gamma, Poisson, normal, inverse Gaussian) can be proved as special cases of GMBM. Keywords: GMBM, GLMs, Classification Ratemaking, Weighted Average.