Small Sample Properties of Robust Analysis of Variance for Fixed and Random Effects Models.

摘要

The major shortcoming of robust methodology in statistical linear models has been a limitation primarily to parameter estimation in fixed effects models. The major competitor, classical least squares methodology, by contrast offers a unified treatment of estimation, testing, and multiple comparisons techniques in a wide range of fixed and random effects models. Research efforts have been directed toward filling the void in robust methods in order to provide a complete alternative to least squares procedures. This report summarizes the results. (Author)