Paper Summary
Share...

Direct link:

Mixed-Effects Location-Scale Models: Methodological and Substantive Considerations

Sun, April 14, 7:45 to 9:15am, Philadelphia Marriott Downtown, Floor: Level 4, Room 402

Abstract

When the assumption of variance homogeneity is violated in multilevel analysis, standard procedure is to estimate robust standard errors. This approach acknowledges the violation but does not fix it. Fitting traditional multilevel models when variance heterogeneity is present may indicate model misspecification. In addition to methodological concerns, modeling level-1 variance heterogeneity may be of substantive interest. Research on teacher ratings of student behaviors typically focus on student differences in overall behavior. However, student differences in behavior variability may be needed. The mixed effects location-scale model is an extension of the standard multilevel model. It can be used to model random variation in within-person behavior. A simulation study is conducted to examine consequences of ignoring random variation in level-1 residual variance.

Author