Tests assessment for multiple comparisons of in vitro gas curves, from the root of the mean square distance

Abstract

In order to find a tool for comparing multiple curves of in vitro gas production to assess ruminants' feeds, the square mean distance between curves is proposed as statistical distance. The performance of four comparison tests of fixed range using this distance: lowest significant distance (LSD), Tukey's honest significant distance (HSD), Scheffé's significant distance (SSD) and Bonferroni-corrected significant distance (BSD) through four experiments simulated by Monte Carlo. The signification level of the tests is also analyzed when the number of treatments or the variation between the simulation parameters increases. It was proved that the Tukey test was stable and limited the inflation, the error type 1, when the number of treatments increases. Besides, it is the most adequate in multiple comparisons of this kind of experiment.