William W. Rozeboom

Linear correlations between sets of variables

While the traditional multiple correlation coefficient appears to be inherently an asymmetrical statistic, it is actually a special case of a more general measure of linear relationship between twosets of variables. Another symmetric generalization of linear correlation is to the total relatednesswithin a set of variables. Both of these developments rest upon thegeneralized variance of a multivariate distribution, which is seen to be the fundamental concept of linear correlational theory.

Sensitivity of a Linear Composite of Predictor Items to Differential Item Weighting.

Many authors have demonstrated for idealized item configurations that equal item weights are often virtually as good for a particular predictive purpose as the item weights that are theoretically optimal. What has not been heretofore clear, however, is what happens to the similarity between weighted and unweighted composites of the same items when the item configurations variance structure is complex.

On Behavioral Theories of Reference

Efforts to bare the psychonomic nature of the semantic reference (representation) relation have been remarkably scanty; in fact, the only contemporary account developed with any care is the one proposed by Osgood. However, not even Osgood has looked deeply at the difficulties that beset any attempt to analyze reference in terms of common effects appropriately shared by a symbol and its significate.

Conditioned generalization, cognitive set, and the structure of human learning

The Ss were given learning trials on which stimulus C was first preceded by signal F and reinforced to response R , followed by reconditioning of C alone to a new response R′ . With high accuracy and no detectable response competition, Ss were able to reproduce either R or R′ in response to F , as well as to C , in compliance with recall instructions.

The Concept of “Memory”

The primary thesis here developed is that virtually nothing in modern research on “memory” has actually dealt with memory at all, for the simple reason that (put oversimply) this research has concerned itself with the retention of associations, whereas memory proper is the recall of beliefs. Related objectives are clarification of the concept of “learning,” and introduction of a methodological distinction between process variables and state variables which has profound importance not merely for analysis of a behavior system’s formal dynamics but also for the practical development of psychological theory.

The Glory of Suboptimal Factor Rotation: Why Local Minima in Analytic Optimization of Simple Structure are More Blessing Than Curse

Yet another facet of factor indeterminacy rears its ugly head: Factor rotation algorithms do not generally find the best solutions of which they are capable. But when enriched with the capacity to conduct repeated searches from random starting positions, a rotation algorithms propensity to converge to optima that are merely local can be fashioned into a seine for catching interpretively provocative rotations of the input factors that might otherwise elude discovery.

The Reliability of a Linear Composite of Nonequivalent Subtests

Traditional formulas for estimating the reliability of a composite test from its internal item statistics are in appropriate to judge the reliability of multiple regres sions and other weighted composites of subtests that are appreciably nonequivalent. Formulas are provided here for the reliability of such a composite given the reliabilities of its component subtests, followed by a comparison of the composites reliability to that of its components. Compositing can easily incur a substan tial loss of reliability, though gains are entirely possi ble as well.

The Cross-Validational Accuracy of Sample Regressions:

Brownes definitive but complex formulas for the cross-validational accuracy of an OSL-estimated regression equation in the random-effects sampling model are here reworked to achieve greater perspicuity and extended to include the fixed-effects sampling model.

The determinacy of common factors in large item domains

Bounds are here developed for the multiple correlation of common factors with the items whose factors they are. It is then easy to see, under broad but not completely general conditions, the circumstances under which an infinite item domain does or does not perfectly determine selected subsets of its common factors.

The validity of the successive intervals method of psychometric scaling

The degree to which scale values computed by the method of successive intervals diverge from theoretically “true” values is seen to be due to three types of error: error due to inequalities in variances of the distributions from which the scale values are computed, error due to non-normality of the distributions, and sampling error. The contribution of each type of error to the total error is evaluated; the latter is seen to be surprisingly small under appropriate conditions. Certain aspects of the formal methodology underlying scaling procedures are also briefly considered.