1997 Volume 24 Issue 1 Pages 112-124
We tested students with MMPI tests on their entrance into medical school, and tried to find the current features of scale scores of the tests as compared with the Japanese reference values obtained more than 25 years before. Since the data consisted of responses for two Japanese language translation versions of the tests we stratified the data by version, sex and calendar year, respectively and calculated means and standard deviations of 14 scale scores in each stratum. By comparing with the mean values of standard population, we found lower Si and higher K scores in both male and female, and significant differences in Mf scores between male and female students. On the other hand, we found neither differences due to temporal trends nor differences between versions of the test. More importantly, we were unable to find differences between successful students and those who had had to repeat a year. Indeed, correct predictions, using a linear discriminant function based upon seven of the scales, were only 60% in male and 64% in female students, respectively. Now, when we used a simple assumption as the diagnostic rule-that no students repeated-then 64% of male and 83% of female students were correctly classified; a result superior to the discriminant function rule. However, by using a different linear discriminant function, with variables selected directly from 550 unit items of MMPI test battery, we were able to attain marginally better separation than the simple rule, but for male students only. In conclusion, it is difficult to obtain a practical prediction rule using MMPI scale scores, even though use of the unit items of MMPI might improve accuracy of predictions.