Regression and Times Series for Actuaries
Regression and Times Series for Actuaries
By Edward W. (Jed) Frees
In January of 1983, the North American actuarial education societies (the Society of Actuaries and the Casualty Actuarial Society) announced that a course based on regression and time series would be part of their basic educational requirements. Since that announcement, a generation of actuaries has been trained in these fundamental applied statistical tools. Sometimes the training has been from an econometrics or business viewpoint that underscores the need to test economic theory. More often, students take two statistics courses, with separate textbooks, in regression and in time series analysis.
I am working on a new book entitled "Regression Modeling with Actuarial and Financial Applications," the first to offer a synthesized introduction to regression and time series for actuaries and other risk managers by introducing contexts that demonstrate practical actuarial applications. This book will emphasize life–long learning by developing statistical tools in a risk management context, providing actuarial applications and introducing more advanced statistical techniques that can be used by actuaries to gain a competitive advantage in situations with complex data. For example, readers will learn about regression and time series using data on the demand for insurance, insurance claims, foreign exchange rates and so on. Although no specific knowledge of actuarial science is presumed, the approach introduces applications where statistical techniques can be used to analyze real data of interest to actuaries. Readers will be exposed to different actuarial practice areas while learning about regression and time series. Moreover, the text will feature many short synopses, or "vignettes," that describe leading edge statistical applications in risk management.
With a foundation in multiple regression and time series, readers will also be introduced to several advanced statistical topics that are particularly relevant to actuarial practice. These topics include the analysis of longitudinal, two–part (frequency/severity) and fat–tailed data. As actuaries become more familiar with related statistical concepts such as data mining and predictive modeling, they learn that a large insurance data set can represent a treasure trove of information to be mined and can yield a strong competitive advantage. The larger the data set, the greater is the need for statistics and statistical methods.
The funding for this project has been provided by the University of Wisconsin (UW). A contract to publish the book with Cambridge University Press; half of the book royalties will go to the UW Hickman Chair of Actuarial Science and half to the Actuarial Foundation. Cambridge University Press has agreed to act as publisher—the paperback version will be less than $50 (U.S.) initially.
The purpose of organizing this material is to educate future and current actuaries on the possibilities of modern statistical techniques, with a focus on regression modeling. Because this project is from academia, help needed is in identifying real–world examples and data—we need compelling illustrations as to how these techniques are used in industry. As the project develops, we will also need reviewers of written materials as well as class testers.
You can learn more about this project at the Web site Research3.bus.wisc.edu. There, you will find many data sets and sample statistical code (in R and SAS) that you can use right away.
Please contact Jed Frees (jfrees@bus.wisc.edu) if you are interested in helping!