The treatment of covariance matrices given by McDonald (1974) can be readily modified to cover hypotheses prescribing zeros and equalities in the correlation matrix rather than the covariance matrix, still with the convenience of the closed-form Least Squares solution and the classical Newton method. (Author/RC)

THIS VOLUME DEALS WITH THE BIVARIATE NORMAL DISTRIBUTION. THE AUTHOR MAKES A DISTINCTION BETWEEN DISTRIBUTION AND DENSITY FROM WHICH HE DEVELOPS THE CONSEQUENCES OF THIS DISTINCTION FOR HYPOTHESIS TESTING. OTHER ENTRIES IN THIS SERIES ARE ED 003 044 AND ED 003 045. (JK)

Results from classical power analysis (Brewer, 1972) suggest that a researcher should not set a=p (when p is less than a) in a posteriori fashion when a study yields statistically significant results because of a resulting decrease in power. The purpose of the present report is to use Bayesian theory in examining the validity of this…

A test is developed for hypotheses about the grand mean in the analysis of variance, using the known relationship between the t distribution and the F distribution with 1 df (degree of freedom) for the numerator. (Author/RC)

It is suggested that Levy's procedure for testing predicted trends in independent correlations should not produce grossly incorrect inferences when one is dealing with the sorts of correlations which are ordinarily encountered in empirical research. (Author/JKS)

Some recent developments for the Shine-Bower single-subject analysis of variance (ANOVA) and the Shine Combined ANOVA are integrated in order to remove the restriction of an even number of trials for the Shine Combined ANOVA. (Author/JKS)

The Koppa coefficient is a measure of association between two variables which have been measured dichotomously. Significance tests for comparing Koppa coefficients from multiple samples are presented. (JKS)

Under mild assumptions, when appropriate elements of a factor loading matrix are specified to be zero, all orthogonally equivalent matrices differ at most by column sign changes. A variety of results are given here for the more complex case in which the specified values are not necessarily zero. (Author/JKS)

A concise method for computing areas under the normal curve using only functions typically found on desk and hand calculators is given. One version of this method will give three-decimal-pace accuracy, and a second, simpler version gives accuracy to two places. (Author/JKS)

The nature of the approximate chi-square test for hypotheses concerning multinomial probabilities is reviewed. Also, a BASIC computer program for calculating the sample size necessary to control for both Type I and Type II errors in chi-square tests for hypotheses concerning multinomial probabilities is described.

The accuracy of the large sample standard error of weighted kappa appropriate to the non-null case was studied by computer simulation for the hypothesis that two independently derived estimates of weighted kappa are equal, and for setting confidence limits around a single value of weighted kappa. (Author/CTM)

Four approximate statistical tests are considered for repeated measurement designs in which observations are multivariate normal with arbitrary variance-covariance matrices. (Author/JKS)

A FORTRAN program, GAMMA, computes Goodman and Kruskal's coefficient of ordinal association, gamma, and Somer's coefficient. The program also provides associated standard errors, standard scores, and probability values. (Author/JKS)

A computer program for testing the association of two nominal variables using Fisher's exact probability test is described. Theoretical corrections according to recent literature are made. Various methods of discovering the source of association at the cellwise level are given. (Author/JKS)