Comparison of Rank Based Tests in Combined Designs
ENAR Biometric Society Spring 2012 Meeting
A simulation study was conducted to investigate the performance of variance statistics for testing differences of effects when k treatments are administered. One constraint under consideration is that data are to be utilized from two experimental designs: a one-way layout and a completely randomized block design. A second constraint is that Normality assumptions are not met. In this study several nonparametric methods based on ranks were considered. Using simulation we compare the performance of tests based on aligned ranks, a rank transformation, and a linear combination of nonparametric tests when sample sizes are not balanced and variances differ under various underlying distributional shapes. The properties of the proposed hypothesis tests are shared.
rank tests, nonparametric
Statistics and Probability
Yvonne Zubovic (2012).
Comparison of Rank Based Tests in Combined Designs. Presented at ENAR Biometric Society Spring 2012 Meeting, Washington, D.C..