At 12:30 p.m. on Thursday, July 25, Dr. Ron Breiger will host a round table entitled "Beyond Regression: New Methods for Analyzing Databases on CBRN Activity" to discuss his and fellow START researcher Lauren Pinson's case-oriented research in regression modeling. The talk will be at START headquarters at 8400 Baltimore Ave.
Many analyses of the use of unconventional weapons rest upon qualitative study of a small number of cases. In contrast, a systematic database, POICN (Profiles of Incidents Involving CBRN by Non-state actors), has recently been compiled by the START Center, University of Maryland. POICN incorporates hundreds of events in which chemical, biological, radiological, or nuclear (CBRN) weapons have been deployed in attacks around the world (1990 to the present), featuring 139 core variables. Regression modeling or some of its generalizations seems an obvious choice for studying outcomes such as the lethality of CBRN events as coded in this systematic new database. However, regression modeling puts the emphasis on variables and on their "average effects," whereas they want to use the variables to learn more about the cases and clusters of cases. Breiger and Pinson apply a new approach that turns regression modeling (and its generalizations) "inside out" in order to present the usual regression coefficients in a new way, as sums across clusters of cases. Their "case-oriented" approach to regression modeling allows them to identify multiple and distinctive profiles that link cases and variables and thus help them to move beyond "one-size-fits-all" prescriptions for countering the risks of unconventional weapons.