Stratifier

A rating factor could alternatively be treated as a covariate and allowed to vary simultaneously with other variables in the model.

Using stratifiers in Generalized Linear Model (GLM) results in a reduction in the amount of data usable for each population subgroup, however, for some factors it remains necessary to segment the data into groups before fitting the GLM to each group. Treating a longevity predictor as a stratifier is generally preferred when:

1. The meaning of other variables differs across the stratifier, greatly complicating the fitting of a GLM; or
2. The shape of mortality with age is very different depending on the value of that stratifier; or
3. The data spans different age ranges for the different values of the stratifier.

Examples of the use of stratifiers are:

• The blue- and white-collar tables developed as part of the Society of Actuaries PRI 2012 study.
• The development of separate tables for 1st lives and 2^nd lives (annuitants and their beneficiaries) in the VitaCurves models.
• The differentiation between male and female tables in most analyses.