After struggling with the problem of representing structure in similarity data for over 20 years, I find that a number of challenging problems still remain to be overcome—even in the simplest case of the analysis of a single symmetric matrix of similarity estimates. At the same time, I am more optimistic than ever that efforts directed toward surmounting the remaining difficulties will reap both methodological and substantive benefits. The methodological benefits that I forsee include both an improved efficiency and a deeper understanding of “discovery” methods of data analysis. And the substantive benefits should follow, through the greater leverage that such methods will provide for the study of complex empirical phenomena—perhaps particularly those characteristic of the human mind.

In connection with a least-squares solution for fitting one matrix, A, to another, B, under optimal choice of a rigid motion and a dilation, Schönemann and Carroll suggested two measures of fit: a raw measure, e, and a refined similarity measure, es, which is symmetric. Both measures share the weakness of depending upon the norm of the target matrix, B, e.g., e(A, kB) ≠ e(A, B) for k ≠ 1. Therefore, both measures are useless for answering questions of the type: “Does A fit B better than A fits C?”. In this note two new measures of fit are suggested which do not depend upon the norms of A and B, which are (0, 1)-bounded, and which, therefore, provide meaningful answers for comparative analyses.

Four prominent oblique transformation techniques—promax, the Harris-Kaiser procedure, biquartimin, and direct oblimin—are examined and compared. Additionally, two newly-developed procedures, falling into the category designated as Case III by Harris and Kaiser [1964], are presented and included in the comparisons. The techniques are compared in light of their freedom from bias in the interfactor correlations, and their ability to yield clear simple structures, over many data sets—some constructed and some “real”—varying widely in terms of number of variables and factors, factorial complexity, and clarity of the hyperplanes. Results are discussed, and implications for practice are noted.

The paper opens with a brief discussion of the problems in testing nonlinear models of attitude change. The regression artifacts produced by unreliability are shown in both the linear and nonlinear case. Classical solutions for the linear case are quickly reviewed. A “new” solution to the linear case is presented and applied to the nonlinear case. It is shown to work well under a broad set of conditions. Regression artifacts in bivariate regression are then discussed. If the predictors are independent, then the univariate correction procedure can be applied to each predictor separately. But if the predictors are correlated, a joint correction procedure must be used. One such procedure is defined and shown to work perfectly in the case of linear regression and reasonably well in a broad set of conditions in which the regression is nonlinear.

Monte Carlo procedures were used to investigate the properties of a nonmetric multidimensional scaling algorithm when used to scale an incomplete matrix of dissimilarities. Various recommendations for users who wish to scale incomplete matrices are made: (a) recovery was found to be satisfactory provided that the “degrees of freedom” ratio exceeded 3.5, irrespective of error level; (b) cyclic designs were found to provide best recovery, although random patterns of deletion performed almost as well; and (c) strongly locally connected designs, specifically overlapping cliques, were generally inferior. These conclusions are based on 837 scaling solutions and are applicable to stimulus sets containing more than 30 objects.

This paper gives a method of estimating the reliability of a test which has been divided into three parts. The parts do not have to satisfy any statistical criteria like parallelism or τ-equivalence. If the parts are homogeneous in content (congeneric), i.e., if their true scores are linearly related and if sample size is large then the method described in this paper will give the precise value of the reliability parameter. If the homogeneity condition is violated then underestimation will typically result. However, the estimate will always be at least as accurate as coefficient α and Guttman's lower bound λ3 when the same data are used. An application to real data is presented by way of illustration. Seven different splits of the same test are analyzed. The new method yields remarkably stable reliability estimates across splits as predicted by the theory. One deviating value can be accounted for by a certain unsuspected peculiarity of the test composition. Both coefficient α and λ3 would not have led to the same discovery.

Let n stimuli be linearly ordered by a physical attribute. Consider an individual's “more than” relation based on perceived attribute values. Assuming that the individual's relation agrees with the linear order and is an interval order, the paper examines the extent to which ordered pairs in the linear order but not in the interval order (due to nondetection of “small” differences) can be correctly identified on the basis of the interval order alone. Different methods of constructing strict weak orders from interval orders are compared by the degree of agreement with the underlying linear order.

Within the context of a contingency table, this note describes the relationship between Freeman's measure of association θ and the asymmetric association measures developed by Somers. The θ coefficient is appropriate for a contingency table in which the levels of one factor are ordered and the levels of the other factor are unordered; the indices defined by Somers are usually used when the levels of both factors are ordered and one is assumed to be the independent factor.

Novick, Lewis and Jackson [1973] propose an interesting Bayesian method for the simultaneous estimation of the probabilities of several binomial distributions. They utilise arc-sine transformations and normal approximations and then refer to the normal theory method of Lindley [1972] which utilises the novel concept of ‘exchangeability’. The posterior estimates for the probabilities shrink the usual proportions towards a central value based on collateral information, and should provide substantial improvements.

The purpose of our comment is to discuss the relative merits of other methods which are available whenever exchangeability of the probabilities is appropriate, The main methods are

(1) The Logistic Approach: Leonard [1972a] uses logistic transformations of the parameters and proposes a simple exact method of estimation which is valid even if some of the cell frequencies are zero. This is because the logistic transformations are to the parameters and not to the observed frequencies. The exact posterior modal estimates replace the usual observed logits in such a way that difficulties involving zeros are removed. The are-sine approximation is of course inaccurate in such cases.