Announcement: SOA releases passing candidate numbers for July 2019 Exam P.

Agenda Day One

Event Home |  Agenda |  Registration Additional Information |  CPD |  Seminar Location |  Presenters

 

Agenda Day One | Agenda Day Two | Agenda Day Three

 

Regression and Generalized Linear Models

Generalized linear models (GLM) are important to a variety of actuarial and insurance applications. They encompass several core techniques like ordinary least-squares (OLS) regression, logistic regression, Poisson regression and Gamma regression.

Day one and a portion of day two will be devoted to this core topic. The morning will be dedicated to lecture-style presentations of core regression and GLM concepts. This theoretical discussion will be complemented by a sequence of textbook-style case studies designed to illustrate theoretical concepts and practical modeling techniques. In the afternoon, the session will transition to a somewhat more realistic, open-ended case study involving modeling health care expenditures. In addition to solidifying the GLM theory presented in the morning, the instructor will continue with data scrubbing, data visualization and exploratory data analysis, and the iterative process of building, criticizing, interpreting and validating models in the afternoon session. 

Wednesday, December 12
7:30 a.m. – 8:00 a.m.
  • Introductions, housekeeping and course overview
  • Orientation: business analytics, data science and their relationships with actuarial science
  • Basic terminology and concepts

Session Coordinator(s)

Facilitator(s)

8:00 a.m. – 9:00 a.m.
  • Regression refresher and R warm-up: the Galton height data
  • Multiple regression concepts and assumptions

Session Coordinator(s)

Facilitator(s)

9:00 a.m. – 11:00 a.m.
  • Sequence of small case studies to illustrate linear modeling concepts and techniques
  • Nested models, analysis of variance
  • Nonlinearities and interactions
  • Cubic spline regression
  • Generalized Additive Models (preview)
  • Session Coordinator(s)

    Facilitator(s)

    11:00 a.m. – 12:00 p.m.
    12:00 p.m. – 1:00 p.m.
    • Theoretical discussion: over-fitting and the bias-variance tradeoff
    • AIC, BIC
    • Out-of-sample validation: the gold standard for model validation
    • Lift curve and gains chart analysis

    Session Coordinator(s)

    Facilitator(s)

    1:00 p.m. – 2:00 p.m.
    • Core components of GLM theory: the exponential family distribution, link functions, analysis of deviance, weights and offsets
    • Overview of special cases of GLM (e.g. OLS, logistic, Poisson, Gamma regression)

    Session Coordinator(s)

    Facilitator(s)

    3:00 p.m. – 4:30 p.m.

    Description of the problem and data, preliminary data exploration

    • Exploratory data analysis
    • Analysis dataset preparation
    • Begin iterative modeling process
    • Interpreting model outputs

    Session Coordinator(s)

    Facilitator(s)