A CONTINUATION STUDY OF USOE PROJECT NUMBER 1493 EXTENDED THE EXAMINATION OF PROBLEMS IN USING MULTIVARIATE SCALING TECHNIQUES FOR THE MEASUREMENT OF INTERESTS. AN EXPERIMENTAL INSTRUMENT, SIMILAR IN FORMAT TO THE SOCIAL RELATIONS SCALE CONSTRUCTED BY GARDNER AND THOMPSON (1956), WAS PREVIOUSLY DEVELOPED. SUBTESTS ON THIS INSTRUMENT (THE INTEREST…

Rotation of canonical components is contrasted with factor rotation and the concept of communalities in canonical analysis is considered. Three possible strategies for rotating canonical components are described: single rotation of all components, separate rotation of each set, and sequential rotation of one set then the other. Constraints,…

A review of model-selection criteria is presented, suggesting their similarities. Some problems treated by hypothesis tests may be more expeditiously treated by the application of model-selection criteria. Multivariate analysis, cluster analysis, and factor analysis are considered. (Author/GDC)

Because ignoring the missing data in an evaluation may lead to results that are questionable, this study investigated the effects of use of four missing data handling techniques on a survey instrument. A questionnaire containing 35 5-point Likert-style questions was completed by 384 respondents. Of these, 166 (43%) questionnaires contained 1 or…

A multivariate post hoc procedure (the Roy-Bose procedure) which allows for the construction of an infinite number of confidence intervals or probability statements concerning the p variables, all of which hold at a 100(1-a) percent level of confidence, is demonstrated and compared to other procedures. (JD)

The internal correlation--a measure of dependency in a set of variables--is discussed and generalized. Applications of the internal correlation coefficient and its generalizations are given for several data-analytic situations. The internal correlation is illustrated and the concept is expanded to a series of additional indices. (TJH)

Use of internal correlation for statistical analysis--proposed by G. W. Joe and J. L. Mendoza (1989)--is discussed. Focus is on the "content" question (what this application can do with the information that statistics contain) and the "eloquence" question (the advantages of this means of encoding information over other means). (TJH)

Multivariate distributions are studied in which all bivariate regressions can be linearized by separate transformation of each of the variables. A two-stage procedure, first scaling the variables optimally and then fitting a simultaneous equations model, is studied in detail. (SLD)

It is shown that goodness-of-fit criteria developed for the evaluation of multivariate structural models can be applied to assist in evaluating the dimensionality of a test consisting of binary items, and correlative methods regularly used in factor analysis can be employed to diagnose causes of misfit. (Author)

Effects of different weighting schemes on selecting the optimal number of observations in multivariate-multifacet generalizability designs are studied when cost constraints are imposed. Comparison of four schemes through simulation indicates that all four produce similar optimal values and that reliability should be similar. (SLD)

This paper clusters and ranks the research performance of 36 Australian universities according to their research performance over the period of 1998-2002. Research performance is measured according to audited numbers of Ph.D. completions and publications, and research grants income (in accordance with rules established by the Department of…

Clerkships would benefit from teachers' improved understanding of the didactic aspects of their task. The purpose of this study is to identify factors that determine the teaching quality of clerkships and to examine the predictive value of these factors for students' global satisfaction. Thus, results would be further reflected to clinical…

The overidentification of structural equation models with latent variables is discussed. The use of two- and three-indicator models is not recommended since such models do not allow a testing of the crucial assumption of unidimensionality among indicators in most cases. Models with four or more indicators may be more sensitive to departures from…

Current practice in structural modeling of variables is limited to means and covariances based on multivariate normality assumptions. This article extends structural equation models to higher order product moments and to non-normal distributions. Areas of possible research are described. (Author/JKS)

Discusses the nature, power, and limitations of four multivariate techniques: factor analysis, multiple analysis of variance, multiple regression, and multiple discriminant analysis. Shows how decision trees assist in interpreting results. (SK)