In response to Webb and Shavelson (EJ 241 567), Nussbaum questions the relevance of the universe score variance as a meaning reliability index and whether it is useful to determine the weights which enter into a composite by means of maximum generalizability. (CM)

Answering Nussbaum's (TM 507 069) criticism, the generalizability coefficient for absolute decisions, the use of the error variance formula, composites of maximum generalizability, and covariance components are discussed as yardsticks of measurement precision with arguments for the use of each procedure to interpret data. (CM)

Known results for testing and estimation problems for patterned means and covariance matrices with explicit linear maximum likelihood estimates are applied to the block compound symmetry problem. An example involving educational testing is provided. (Author/JKS)

The MMPI results obtained from 245 adolescent males referred to the evaluation unit of a Juvenile Court were submitted to a multivariate classification system. By correlating individual subject profiles with the modal profiles, six membership groups were formed. No relationship was found between group membership and age or race. (Author)

A particular strategy for investigating effects from a multivariate analysis of variance (MANOVA) is proposed. The strategy involves multiple two-group multivariate analyses. The analysis strategy is described in detail and illustrated with real data sets. (Author/JKS)

Canonical correlation analysis is considered to be a general model for bivariate and multivariate statistical methods. Some problems involving assumptions and statistical tests for parameters exist for social science data. A resolution for these problems is presented by treating canonical correlation as a special case of linear structural…

A distinction is drawn between redundancy measurement and the measurement of multivariate association between two sets of variables. Several measures of multivariate association between two sets of variables are examined. (Author/JKS)

It is demonstrated that, in the presence of population information, a linear combination of true scores can be estimated more efficiently than by the same linear combination of the observed scores. Three criteria for optimality are discussed, but they yield the same solution, described as a multivariate shrinkage estimator. (Author/SLD)

Use of internal correlation for statistical analysis--as proposed by G. W. Joe and J. L. Mendoza (1989)--is discussed. Use of the bootstrap technique to deal with the distributional problem is questioned. Joe and Mendoza attempt the interpretation of the two linear composites that produce the largest internal correlation. (TJH)

The use of internal correlation for statistical analysis, as proposed by G. W. Joe and J. L. Mendoza (1989), is discussed. The suggestion of using bootstrapping is received well. Applications to collinearity are suggested. (TJH)

The source of an event-related brain potential (ERP) is estimated from multivariate measures of ERP on the head under several mathematical and physical constraints on the parameters of the source model. Statistical aspects of estimation are discussed, and new tests are proposed. (SLD)

Interpretations of relative variable importance in multivariate analysis of variance are discussed, with attention to (1) latent construct definition; (2) linear discriminant function scores; and (3) grouping variable effects. Two numerical ranking methods are proposed and compared by the bootstrap approach using two real data sets. (SLD)

A generalized measure of association and an associated test of significance are presented for nominal independent variables in which any number or combination of interval, ordinal, or nominal dependent variables can be analyzed. A permutation test of significance is provided for the new measure. (SLD)

Examination of the stochastics of moderated multiple regression (MMR) reveals that MMR is an appropriate technique when predictors are fixed variables and the distribution of errors is normal but is not appropriate when predictors are random variables and the joint distribution of criterion and predictor variables is multivariate normal. (SLD)

Within the multivariate normality framework, a formula is provided for computation of the criterion-related validity of composite scores based on the highest (or lowest) of several equivalent measures. This partial composite score has more validity than each single observation, but less validity than a composite based on all observations. (SLD)