This Monte Carlo study was designed to investigate empirically the efficacy of three statistical strategies for estimating the relationships in extreme group designs in terms of estimation of correlation, power and mean square error for the correlation. Alf and Abrahams's "covariance information statistic" proved to be the best strategy.…

To modify least squares estimators so that outliers arising from heavy-tailed distributions will have less influence, a robust coefficient of kurtosis, which avoids the quartic effect, is described. Aspects of Pearson's and Fisher's kurtosis coefficients are compared in the procedure. (Author/CM)

Equations for estimating the value of the multiple correlation coefficient in the population underlying a sample and the value of the population validity coefficient of a sample regression equation were investigated. Results indicated that cross-validation may no longer be necessary for certain purposes. (Author/MH)

"Composition analysis" refers to the mathematics of how component measures combine to produce composite measures based on Bookstein's functions. Composition analysis is applied to the different ways group members might work together such that their individual measures can be combined mathematically to calculate an expected measure for…

Almost all modern rotation of factor loadings is based on optimizing a criterion, for example, the quartimax criterion for quartimax rotation. Recent advancements in numerical methods have led to general orthogonal and oblique algorithms for optimizing essentially any rotation criterion. All that is required for a specific application is a…

Modeling data behavior is differentiated from statistically determining a best-fit regression function, as it allows students to develop functions that hold certain attributes in respect to a set of data. Three techniques for developing polynomial models are demonstrated that afford students optional methods for determining appropriate polynomial…

Formulas for estimating cross-validated hit-rates, the number of correct classifications into an a priori grouping structure, were examined. The following mathematical formulas were compared: McLachlan's formula estimator, two Snappin and Knoke smoothed formula estimators, and the analytic leave-one-out estimator. The R method was included as a…

The Welch-James procedure may be used to test hypothesis on means, when independent samples from populations with heterogenous variances are available. Summation formulas for the Welch-James procedure are presented for the 2x2 design. Matrix formulas that permit routine application of the procedure to crossed factorial designs are presented.…

Different indices for the internal consistency, reproducibility, or homogeneity of a test are based upon highly similar conceptual frameworks. Illustrations are presented to demonstrate how the maximum and minimum values of KR20 are influenced by test difficulty and the shape of the distribution of test scores. (Author/CTM)

Describes a procedure for information retrieval testing that is based on the comparison of statistically independent methods of retrieval applied to the same database. The probability ranking principle is discussed, the statistical meaning of relevance is examined, and the methodology is illustrated on a large database of MEDLINE records. (19…

Describes a study in which several different sets of hypertext links are inserted by different people in full-text documents. The degree of similarity between the sets is measured using coefficients and topological indices. As in comparable studies of inter-indexer consistency, the sets of links used by different people showed little similarity.…

Studies binomial, negative binomial, and gamma regression models and gives a detailed description of inference procedures based on them. The process of model fitting and evaluation is illustrated by examples referring to the determination of endpoints in acute and chronic toxicity tests. (17 references) (Author/MDH)

There are a variety of formulas for reducing the positive bias which occurs in estimating R squared in multiple regression or correlation equations. Five different formulas are evaluated in a Monte Carlo study, and recommendations are made. (JKS)

The problem of monitoring trends for changes at unknown times is considered. Statistics that permit one to focus high power on a segment of the monitored period are studied. Numerical procedures are developed to compute the null distribution of these statistics. (Author)

The merits of double exponential smoothing are discussed relative to other types of pattern-based enrollment forecasting methods. The difficulties associated with selecting an appropriate weight factor are discussed, and their potential effects on prediction results are illustrated. Two methods for objectively selecting the "best" weight…