Panel Paper: Comparing Boolian and Logit Analysis: An Application to Rare Events Data

Thursday, November 6, 2014 : 1:20 PM
Isleta (Convention Center)

*Names in bold indicate Presenter

Kendra Koivu and Chris Butler, University of New Mexico

Methodological diversity is praised in some circles for providing a larger set of tools to the researcher but it is possible that different tools provide different answers to the same question. How do our methodological choices affect the answers we get? One area of methodological diversity that provides some similarity to large-N statistical analysis is Boolean analysis. We analyze data from both approaches. We start by examining simulated rare-events data sets following both logistic and Boolean logics, analyzing them separately, and then comparing results. When the data generation process is known, we find that both statistical and Boolean approaches provide largely the same results. When the data generation process is not known, however, the results are dependent on the choices researchers make. We then examine civil-war onset using both approaches. We find that large-N, rare-events data sets present challenges to both statistical and Boolean approaches. While these challenges have been largely resolved in statistical studies, Boolean approaches have yet to establish best practices. We offer suggestions from our study regarding how Boolean analysis can tackle large-N, rare-events data sets. We also summarize the findings of Boolean analysis regarding civil-war onset and compare these findings with extant statistical findings in the literature.