Analytics You Should Know but Don’t

Analytics You Should Know but Don't

By Steve Craighead

On Oct. 15, 2007 at the SOA Annual Meeting in Washington, DC, the E&R Section sponsored a predictive modeling session called "Analytics You Should Know, but Don't." The session was moderated by Steve Craighead, ASA, MAAA, who also was a panelist. The other speakers were Chris Kryder, MD, MBA, chairman and CEO of D2Hawkeye Incorporated, and Dave Wilson, FSA, FCIA, MAAA, president of the Windsor Strategy Partners and Margrét Viborg Bjarnadóttir from the MIT Operations Research Center.

Craighead outlined the use of the open source R language to the audience and discussed several metrics that available in the R contributed Packages. He also briefly discussed his results on the estimation of PBA reserves and capital utilizing a combination of scenario reduction and predictive modeling.

Dr. Kryder outlined the history of his company and discussed how his company has developed several tools using predictive modeling to aid in health underwriting and case management.

Wilson discussed a Diabetic Population business case that demonstrated how the D2Hawkeye tool RI classified 6,306 diabetic members into seven separate cost categories. He also discussed the effectiveness of the tool in the prediction of future costs as well.

Bjarnadóttir described the MIT Predictive Project which was a collaborative work between MIT and D2Hawkeye. She also spent some time on various discussions of the statistics that underpinned the software.

The entire session is available in an electronic format.