A new estimator of the actual error rate for the two-group discriminant problem is presented. The new estimator is based on the concept of a shrunken generalized distance. This new estimator is compared favorably to two modified distance estimators and an estimator developed by M. Okamato. Four figures are included. (Author/TJH)

Although comments of D.J. Nordlund and R. Nagel are welcomed, their arguments are not sufficient to accept the recommendation of using total variance estimates to standardize canonical discriminant function coefficients. If standardized coefficients are used to help interpret a discriminant analysis, pooled within-group variance estimates should…

Fisher's two-group discriminant function has been generalized in two different ways for the case of three or more groups, leading to confusion in the literature. The precise functional relation between the two functions is derived, and the interpretation of the two functions is discussed. An example is provided. (Author/CTM)

Reasons/methods for standardizing discriminant function coefficients are reviewed, and use of the pooled within-groups variance estimate is examined. Use of total or pooled within-groups estimates can be justified only on purely mathematical grounds. Total variance estimates offer a more consistent and parsimonious approach than do pooled…

The relationship between quantitative predictor variables and the probability of occurrence of one or more levels of a qualitative criterion variable can be analyzed by quantal response techniques. This paper presents and discusses two quantal response models, comparing them to multiple linear regression and discriminant analysis. (Author/JKS)

Standardization procedures in discriminant analysis are discussed. Three leading software packages--SPSSX, BMDP, and SAS--are compared in terms of their calculations of unstandardized and standardized discriminant coefficients. Estimation procedures are described for each. Arguments are presented for within-group, rather than total, variance…

Data on all applicants to U.S. medical schools were analyzed to find how applicant characteristics affected the probability of admission, particularly the function of state of residence separately for majority and minority group applicants. Discriminant analysis and empirical Bayes methods were used, and the differences were discussed. (Author/CTM)

A multilevel variance component analysis from the pilot year of pretesting an instrument--the GENED--designed to present information about general education outcomes is presented, using data from about 11,000 college students. The analysis addresses the discriminant validity of the subtests and statistical issues in test construction. (SLD)