January 19, 2021
The CAS Institute is pleased to offer a new instructor-moderated online study group to help candidates prepare for iCAS Exam 3: Predictive Modeling – Methods and Techniques. The exam has been rescheduled to Thursday, June 10, in order to coordinate with the study group schedule below. Dorothy Andrews, ASA, MAAA, CSPA* will be available online to answer questions submitted on the specified Learning Objectives on the assignment-based schedule below.
Ideally students will work through a topic by the week prior to instructor availability. Posting specific topic-related questions to the online platform will allow the instructor to provide timely answers.
There is no charge to participate in the study group. To register, visit https://bit.ly/3i8xn1c.
| Week | Week of | Topic |
| Classical Models & Diagnostics | ||
| 1 | 2/1 | · Types of Data, Missing and Incomplete Data |
| 2 | 2/7 | · Linear Model Diagnostics |
| 3 | 2/15 | · Classical Models – Generalized Linear Models and Their Diagnostics |
| 4 | 2/22 | · Hierarchical Models, Including Linear Mixed Models, and Buhlmann Credibility |
| Machine Learning Methods | ||
| 5 | 3/1 | · Validation Evaluation: Goodness of Fit Metrics, Bootstapping, Holdout vs. Cross-Validation and Tuning Parameters |
| 6 | 3/8 | · Evaluation: Goodness of Fit Metrics, Bootstrapping, Bias-Variance Tradeoff, and Presentation of Results |
| 7 | 3/15 | · Classification Models and Special Considerations |
| 8 | 3/22 | · Shrinkage and Feature Selection Methods |
| 9 | 3/29 | · Non-Linear Effects and Additive Models |
| 10 | 4/5 | · Single Trees |
| 11 | 4/12 | · Ensemble Methods, Random Forests, and Boosting |
| 12 | 4/19 | · Principal Components Analysis and Unsupervised Learning |
| 13 | 4/26 | · Application Specific Methods: Association Models |
| 14 | 5/3 | · Application Specific Methods: Fraud Detection |
| Causal Inference | ||
| 15 | 5/10 | · Experimental Design |
| 16 | 5/17 | · Experimental Methods |
| 17 | 5/24 | · Causal Inference from Observational Data |
*Study Group Moderator/Instructor Dorothy Andrews is the Chair of the American Academy of Actuaries’ Data Science and Analytics Committee and the Chief Behavioral Data Scientist for the Actuarial and Analytics Consortium. She has more than 25 years of actuarial and predictive modeling experience with life insurance companies, property and casualty insurance companies, reinsurance companies, international consulting firms, and government agencies. She is adjunct faculty at the University of North Carolina at Charlotte, where she teaches predictive analytics. She taught in the actuarial science master’s degree program at Boston University for 7 years and won the outstanding teaching award in 2002.
An Associate of the Society of Actuaries, a Member of the American Academy of Actuaries, a Fellow of the Conference of Consulting Actuaries, and a Certified Specialist in Predictive Analytics (CSPA), Dorothy is currently working on a Ph.D. in Media Psychology.
