ABSTRACT
Simplex analysis applies only to a limited class of problems, namely the class in which all of the variables can be arranged in an ascending or descending order of complexity and are otherwise similar. The procedure consists merely in fitting the simplex model to the correlation matrix of the ordered variables, and then checking to see how well the model fits. Simplex analysis applies most clearly to variables which have a natural order that might be due to increasing or decreasing complexity, such as successive trials in a learning study, successive measurements in a growth study, and the like. Large tables of intercorrelations seldom exhibit the superdiagonal pattern or the simplex form. Sampling errors disturb the fit of a simplex only randomly, but unique factors disturb it systematically as well. In fitting a quasi-simplex, the investigator should first order the variables according to his best judgment of their increasing or decreasing complexity.
