Zhiyong
Mediation models have been extensively studied and used in psychological research. Among the methods testing mediation effects, the bootstrap resampling method has been shown more effective than the large sample method in small sample research. In the current study, a modified bootstrap method is proposed which is based on resampling residual errors of the mediation models in the framework of path analysis. The proposed method is then compared with the large sample method and bootstrapping raw data method in different conditions. The results show that the bootstrapping error method has better coverage probability and is more efficient than the bootstrapping raw data method when the residual errors are homogeneous. In the heteroscedastic case, the bootstrapping raw data method performs best among three methods regardless of sample size. A C++ program is provided to implement all three methods with the modified Brown-Forsythe statistic to test the homogeneity of residual errors.
Click the link to view the schedules of DADA talks. http://www.cdhrm.org/article.php?articleid=83