contrast coding
in analysis of variance and multiple regression, a method of assigning weights to values so that specific comparisons of interest can be made. To ensure the statistical independence of the particular quantities being evaluated, the process requires that the weights in each comparison sum to 0. For example, consider a researcher investigating four different methods for teaching course content. If he or she wanted to compare the average performance of the first group to the average performance of the remaining three groups, the investigator would apply contrast weights of +3, −1, −1, and −1 to the respective group means. If, however, he or she wanted to compare the average performance of the first group to the average performance of the fourth group, the investigator instead would apply contrast weights of +1, 0, 0, and −1 to the respective group means. Also called orthogonal coding. Compare dummy variable coding;
effect coding.