Vortrag von

• Tenko Raykov, Ph. D. (Michigan State University)

This workshop presents an introduction to multilevel modeling with a focus on scale reliability evaluation in settings with clustering effect. Initially the need for hierarchical modeling is discussed and the disadvantages of "classical" statistical methods when analyzed data is affected by nesting effects are explicated. Traditionally employed methods for aggregation and disaggregation are also criticized.

A brief example is used to demonstrate the value of hierarchical linear modeling on a data set stemming from an industrial organizational study. The intra-class correlation coefficient is then discussed and illustrated on empirical data. A widely applicable method for estimation of scale reliability with clustering effects is then focused on and exemplified on data. The approach provides a readily applicable means for testing scale unidimensionality as well as interval estimation in settings with nesting effects.

Throughout the workshop multiple uses are made of the popular statistical analysis and modeling software Stata and Mplus.