|
Abstract:
|
Two statistical software packages, HLM and SUDAAN, can be used to analyze clustered data. It has been suggested that SUDAAN is more robust than HLM to departures from normal homoscedastic errors. We investigated this question directly with a series of simulations. Specifically, we examined robustness in the context of a randomized complete block design, where each "plot" is a small group of children at the same nursery school and a series of measurements of each child are made. We constructed a series of superpopulations in which the standard assumptions of hierarchical (mixed effects) linear models were violated. In addition to testing HLM and SUDAAN, we tested SAS PROC MIXED, and a semi-parametric analysis of variance along the lines discussed by Rosenbaum (2002).
|
