It is common in regression discontinuity analysis to control for high order (third, fourth, or higher) polynomials of the forcing variable. We argue that estimators for causal effects based on such methods can be misleading, and we recommend researchers do not use them, and instead use estimators based on local linear or quadratic polynomials or…

We focus on the problem of ignoring statistical independence. A binomial experiment is used to determine whether judges could match, based on looks alone, dogs to their owners. The experimental design introduces dependencies such that the probability of a given judge correctly matching a dog and an owner changes from trial to trial. We show how…

Designs for sequential sampling procedures that adapt to cumulative information are discussed. A familiar illustration is the play-the-winner rule in which there are two treatments; after a random start, the same treatment is continued as long as each successive subject registers a success. When a failure occurs, the other treatment is used until…

A linear structural model for comparing quasi-Markov models across populations is demonstrated. A confirmatory factor analysis formulation of the simplex model is also developed for between group comparisons. A variety of possible applications of this approach are suggested. (Author)

The validity conditions for univariate repeated measures designs are described. Attention is focused on the sphericity (equality of variance) requirement. It is recommended that separate rather than pooled error term procedures be routinely used to test a priori hypotheses. (Author/JKS)

Linear equating methods for the common-item non-equivalent populations design were compared when true-score correlation between the test and anchor was less than unity. Scores from two groups of approximately 300 examinees illustrated three methods: (1) the Tucker equating method; (2) the Angoff-Levine method; and (3) the Congeneric-Levine method.…

Analysis of covariance (ANCOVA) is a data analysis method that is often used to control extraneous sources of variation in non-equivalent group designs. It is commonly believed that as long as the covariate is highly correlated with the dependent variable there is nothing to lose in using ANCOVA, even in non-randomized studies. This paper examines…

Glass's estimator of effect size, the sample mean difference divided by the sample standard deviation, is studied in the context of an explicit statistical model. The exact distribution of Glass's estimator is obtained and the estimator is shown to have a small sample bias. Alternatives are proposed and discussed. (Author/JKS)

A brief, practical overview of "design effects" (DEFFs) is presented for users of the results of sample surveys. The overview is intended to help such users to determine how and when to use DEFFs and to compute them correctly. DEFFs are needed only for inferential statistics, not for descriptive statistics. When the selections for…

This article reviews some of the less well-known hypothesis tests for variance, how they are employed, and how the results may be interpreted. Tests include testing for a single variance and the T-test for correlated variances. (CT)

Browne's 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. (Author)

In 1955, R. Levine introduced two linear equating procedures for the common-item non-equivalent populations design. His procedures make the same assumptions about true scores; they differ in terms of the nature of the equating function used. In this paper, two parameterizations of a classical congeneric model are introduced to model the variables…

The use of analysis of variance was examined under the assumption that the treatment had been randomly assigned to students, when in fact, the class had been the unit. Data support the idea that if one can randomly assign treatments to intact classes, consideration should certainly be given to doing so. (Author/GK)

A methodology for determining the optimal number of observations to use in a measurement design when resource constraints are imposed is presented. Two- and three-facet designs are outlined. Parallel closed form formulae can easily be determined for other designs. (TJH)

The routine epsilon approximate test in repeated measures designs when the condition of circularity is unfulfilled uses an erroneous formula in the case of two or more groups. Because this may lead to underestimation of the deviation from circularity when the subject number is small, a correction is proposed. (Author/SLD)