A Note Regarding Evaluation of Multiple Regression Models

Econometric multiple regression models are now commonplace aids to understanding variables affecting the insurance industry. For actuaries and other corporate management personnel to utilize these models to fullest advantage, it is necessary to be familiar with important regression statistics and to be able to critically evaluate model structure. This paper discusses statistics for determining the strength or validity of a model. Special emphasis is given to the definition of the R’ statistic and its relationship to the R’ and F statistics. Exclusion of constants from causal models is recommended. Reasons for modeling change in dependent variable rather than level of the variable are considered.