The importance of the experimental error

Abstract

The statistical analysis is used to test hypotheses due to the presence of the effects of not controlled
factors that cause variation, called experimental error. These effects are randomized among the units that received treatments, they cannot be known individually and they have direct interference in the tests of hypotheses and in the procedures for comparisons of means. The objective of this review is to discuss the importance of the
experimental error in F test and least significant difference (LSD) used in the procedures of multiple comparisons of means, using as example a competition assay of common beans cultivars. LSD increases with the increase of the mean square residues. When the coefficient of variation increases from 5% to 20%, LSD increases four times
in the tests of Scheffé and Tukey and five times for Dunnett and Duncan tests. For a coefficient of variation of 20%, LSD is greater than 50% of the mean of cultivars for Scheffé and Tukey tests. With a coefficient of variation of 35% the LSD between two treatments is higher than the mean experiment when the Scheffé test is used.