An aligned rank transform test in randomized complete block designs

Abstract EN

This article provides an insight into the power and Type I error of alternative test for experimental effects in the randomized complete block designs (RCBD).

An aligned rank transform test (AR) provides a useful alternative method of analysis when the Gaussian assumptions are not met.

The AR test consists of aligning the data by removing blocks effect, then ranking them and performing the F test on the ranks.

The AR test can be implemented with existing statistical packages and provides a fix-up for the rank transform procedure.

Simulation results suggest that the test is valid for small and moderate treatment numbers which usually are used in practice when error distributions are symmetric or moderately skewed.

The procedure has advantages over F test in the standard analysis of variance when error distributions are heavy tailed.