Some Consequences When the Assumptions for the Analysis of Variance are not Satisfied

Publisher Website
Google Scholar
Abstract
The consequences when each of the assumptions, which underlie the analysis of variance, are not satisfied are discussed. In general, the factors liable to cause the most severe disturbances are extreme skewness, the presence of gross errors, anomalous behavior of certain treatments in parts of the experiment, marked departures from the additive relationship, and changes in the error variance, either related to the mean or to certain treatments or parts of the expt. The principal methods for an improved analysis are the omission of certain observations, treatments, or replicates; subdivision of the error variance; and transformation to another scale before analysis. These methods are illustrated by several examples.
Keywords

This publication has 2 references indexed in Scilit: