Simple Bayesian Inference for Qualitative Political Research

In political science and related disciplines in the social andbehavioral sciences, there exists an unfortunate de facto dividebetween qualitative and quantitative empirical research. Sometimesthis divide is purely a function of training and disciplinarysocialization, but often it reflects a valid dispute over thephilosophical foundations of inquiry. I argue here that theBayesian approach to quantitative empirical modeling is an amenablestarting point for building a rapprochement between qualitativeand quantitative research, and I introduce as an example a straightforwardmodel that allows for the Bayesian estimation of the differencebetween means of very small samples with unknown and possiblyunequal variances. I then extend this approach to consider nonnormalvariates, informative priors, and a multivariate test of thedifference of means useful for the researcher who is interestedin determining whether two small samples are different on severaldimensions simultaneously.     

Comparability of the MMPI-2 F(p) and F scales and the SIRS in clinical use with suspected malingerers

Relationships among Structured Interview of Reported Symptoms (SIRS) scores and Minnesota Multiphasic Personality Inventory--2 (MMPI-2) F(p) and F scores were examined for 63 suspected malingerers evaluated at either of two psychiatric facilities. Despite differences between facilities in terms of seriousness of subjects' offenses, mean scores on the malingering tests were similar. Cutting scores for F(p) and F resulting in substantial correspondence between these scales and the SIRS were derived. Use of the cut score for F(p) proposed by Arbisi and Ben-Porath (1995) resulted in less agreement with the SIRS than did a lower cut score. No substantial difference between F(p) and F in each scale's overall agreement with the SIRS was observed. A principal components analysis of the SIRS primary scales produced two factors, interpreted as Overreporting of Symptoms and Implausible Symptoms. F(p) was observed to correlate significantly with Implausible Symptoms but not with Overreporting of Symptoms; F was significantly correlated with both factors.