Bayesian Regressions for Cross-Validation: An Application

Summary

It should be noticed that the traditional correlation analysis for investigating the validity of a psychological test is often insufficient because the correlation coefficient is usually obtained by maximizing the likelihood, which is derived from the given data set. Furthermore, a simple correlation makes no sense when the criterion variable is not completely the same among the several groups of subjects. In this paper, a Bayesian cross-validation using hierarchical regression model is considered. In order to investigate the predictive effectiveness of a test, the method of deriving the cross-validity index is considered. A practical application to the Numerical Reasoning Ability Test for Myanmar high school students is also shown, which contains the scores of school subject tests as the criterion variable. Then, it is found that the cross-validity index is useful because it indicates the predictive effectiveness of the NRT to the various school subject tests.