Multivariate Behavioral Research

Gliner, Gail; Goldman, Susan R.; Hubert, Lawrence J.

doi: 10.1207/s15327906mbr1801_1pmid: 26764552

Exploratory multidimensional scaling and confirmatory nonparametric procedures (Hubert and Levin, 1976) were used to represent data from similarity rating and sorting tasks performed on nine animal names. Confirmatory procedures demonstrated that the organization of the data from the two tasks was similar. Analyses of data from sorting tasks performed after reading two stories with the nine animals as main characters (Bisanz, LaParte, Vesonder, and Voss, 1978) suggested a change from pre-reading organization that was similar to the organization of the characters intended by the authors in one of the two stories. One of the two dimensions used to write the second story appeared not to be salient to the readers.

Holden, Ronald R.; Reddon, John R.; Jackson, Douglas N.; Helmes, Edward

doi: 10.1207/s15327906mbr1801_2pmid: 26764553

Three samples consisting of normal adults (n=182), psychiatric patients (n=352) and high school students (n=1,485) were used to evaluate the empirical item structure of the Basic Personality Inventory. Using a principal components model with a confirmatory rotation, 96, 99, and 98 percent of the items loaded appropriately on their respective scales. No items were found to load inappropriately across all three samples, 219 of 220 items loaded appropriately across two samples, and 205 of 220 items loaded appropriately across all three samples. The empirical item structure was found to be congruent with the hypothesized structure, with keyed items showing much higher loadings than non-keyed items. All results indicated a substantial improvement over that expected by chance. These results were interpreted as providing evidence for the efficiency of the construct-oriented heuristic.

Lee, Raymond; McCabe, Dennis J.; Graham, William K.

doi: 10.1207/s15327906mbr1801_3pmid: 26764554

This study investigated multivariate relationships between task characteristics, measured by the Job Diagnostic Survey, and satisfaction with work outcomes, measured by the Triple Audit Opinion Survey, for 1,972 workers in the public sector. The obtained canonical correlations were high and stable across independent samples of public employees as indicated by a triple cross-validation design. The results support the predictions derived from the Hackman and Oldham model that job characteristics are more important for feelings of intrinsic job satisfaction than feelings of extrinsic job satisfaction, and that job characteristic - job satisfaction relationships are higher for high self-actualization need strength employees than for low self-actualization need strength employees. It was concluded that previous job characteristic - job satisfaction relationships established for private sector employees can be generalized to public sector employees.

Jensen, Robert E.

doi: 10.1207/s15327906mbr1801_4pmid: 26764555

An eigenvector approach to scaling of choice alternatives based upon multiple attributes evaluated on a relative basis by paired comparisons is reviewed. It is shown that neither compensatory aggregation (composition) nor noncompensatory models may be consistent or appropriate for an individual's utility. An alternative approach, where the individual (judge) evaluates all attributes simultaneously, and thereby, performs a subjective aggregation (composition) across all attributes simultaneously, is proposed. Eigenvector scaling may then be used to "prioritize" choice alternatives based upon aggregated pairwise comparisons.

Dong, Hei-Ki

doi: 10.1207/s15327906mbr1801_5pmid: 26764556

In many instances investigators are hesitant to use "too many" objects or stimuli in a multidimensional scaling study using the method of complete triads. One of the major reasons for this is that such use is postulated to lead to subject fatigue/boredom effects or practice effects which may produce unreliability in the judgments. In the present study an experimental design was constructed in an attempt (a) to investigate the presence of subject fatigue/boredom effects or practice effects and (b) to assess the influence of these effects on the configurations and distances resulting from multidimensional scaling of nations. The results showed that the postulated effects were present, but they did not influence the multidimensional scaling solutions.

Tate, Richard L.

doi: 10.1207/s15327906mbr1801_6pmid: 26764557

The versatility of discriminant analysis as a technique to obtain more parsimonious descriptions in multivariate studies has not been fully exploited. Although the technique has been commonly proposed and used for multivariate analysis of variance studies, there have been few applications for other models and effects of current interest in the behavioral sciences. This article briefly reviews generalized discriminant analysis as a descriptive technique associated with the multivariate general linear model, and then presents results for several example analyses. The examples, based on real data, include contextual effects analyses using multilevel models and an aptitude-treatment-interaction analysis. It is argued that discriminant analysis has the most value as a descriptive technique when used in the spirit of an "external factor analysis."

Cliff, Norman

doi: 10.1207/s15327906mbr1801_7pmid: 26764558

Literal acceptance of the results of fitting "causal" models to correlational data can lead to conclusions that are of questionable value. The long-established principles of scientific inference must still be applied. In particular, the possible influence of variables that are not observed must be considered; the well-known difference between correlation and causation is still relevant, even when variables are separated in time; the distinction between measured variables and their theoretical counterparts still exists; and ex post facto analyses are not tests of models. There seems to be some danger of overlooking these principles when complex computer programs are used to analyze. correlational data, even though these new methods provide great increases in the rigor with which correlational data can be analyzed.