The Use of Adjusting Factors in the Analysis of Data with Disproportionate Subclass Numbers

Publisher Website
Google Scholar
Abstract
Coding by means of the equation: [image]a[long dash][image]1+ [image] = [image]n, (where [image]a is the unadjusted mean of the ith. subclass in the ith row or column, Aij is the corresponding adjusted or coded mean, Xj is the mean of the jth row or column, and X is the grand mean) will eliminate the variance attributable to either of the border effects in a 2-way table designed for analysis of covariance. When applied to a table which is non-orthogonal as a result of disproportionate subclass numbers the removed sum of squares will be too large because of interaction. This deficiency may be recovered by successively readjusting alternately for intercolumn and interrow differences, finally yielding the same result as the method of Brandt and Yates involving the least-squares fitting of constants. When the latter method is applicable, Patterson''s method should reduce the labor of computation. One orthogonal and one . non-orthogonal example are fully worked out.

This publication has 1 reference indexed in Scilit: