One common barrier to employing GLM modeling is the lack of sufficient data. This limitation often comes from the combination of the considerations in ASOP No. 25 “Credibility Procedures” and the fact that GLMs assign full credibility to the data. Explore the possibilities of credibility-based regression by expanding the intended purpose of modeling analysis from producing a single model as output to producing data insights, credible differences, and a range of reasonable models.
Come examine practical use cases of penalized regression on large, medium, and small datasets and supporting considerations to make these analyses actuarially sound. Participants are expected to complete a rubric to identify possible additional use cases of penalized regression models on current or future projects at their company or in their role.