Prediction for a number of criteria from a set of predictor variables in a system of regression equations is studied with the possibilities of linear transformations applied to both the criterion and predictor variables. Predictive composites representing a battery of predictor variables provide identical estimates of criterion scores as do the…

This paper discusses how full model dummy variables can be used with partial correlation or multiple regression procedures to compute matrices of pooled within-group correlations. (Author/CTM)

The t ratio used in testing the difference between two independent regression coefficients is generalized to the multivariate case of testing the difference between two vectors of regression coefficients. This is particularly useful in determining which of two variables best predicts a number of criterion variables. (Author/JKS)

Data were simulated to conform to covariance patterns taken from personnel selection literature. Incomplete data matrices were treated by four methods. Treated matrices were subjected to multiple regression analyses. Resulting regression equations were compared to equations from original, complete data. Results supported using covariate…

The optimum weighting of variables to predict a dependent-criterion variable is an important problem in nearly all of the social and natural sciences. Although the predominant method, multiple regression analysis (MR), yields optimum weights for the sample at hand, these weights are not generally optimum in the population from which the sample was…

The results of an algorithm which is designed to take a set of commonality coefficients, either real or manipulated, and, if possible, produce one or more sets of regressor correlations that are consistent with them are examined. A number of different ways of resolving the higher order commonality values into their lower orders were tried and the…

Commonality analysis is an attempt to understand the relative predictive power of the regressor variables, both individually and in combination. The squared multiple correlation is broken up into elements assigned to each individual regressor and to each possible combination of regressors. The elements have the property that the appropriate sums…

The use of characteristic coding (dummy coding) is made in showing solutions to four multivariate problems using canonical analysis. The canonical variates can be themselves analyzed by the use of multiple linear regression. When the canonical variates are used as criteria in a multiple linear regression, the R2 values are equal to 0, where 0 is…

Three methods of estimating a model of school effects are compared: ordinary least squares; an approach based on the analysis of covariance; and, a residualized input-output approach. Results are presented using a matrix algebra formulation, and advantages of the first two methods are considered. (Author/GK)

A general linear model (GLM), using least-squares techniques, was used to develop a criterion measure to replace freshman year grade point average (GPA) in college admission predictive validity studies. Problems with the use of GPA include those associated with the combination of grades from different courses and disciplines into a single measure,…

The effects of the violation of the assumption of normality coupled with the condition of multicollinearity upon the outcome of testing the hypothesis Beta equals zero in the two-predictor regression equation is investigated. A monte carlo approach was utilized in which three differenct distributions were sampled for two sample sizes over…

An "undesigned" experiment is one in which the predictor variables are correlated, either due to a failure to complete a design or because the investigator was unable to select or control relevant experimental conditions. The traditional method of analyzing this class of experiment--multiple regression analysis based on a least squares…