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…

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…

In large-scale surveys, it is no longer uncommon for repeated measurements to be obtained from respondents and analyses performed to gauge the magnitiude of nonsampling errors. This is particularly true for periodic surveys and longitudinal surveys where a very large investment in data collection is made. This technical note, aimed at the analysis…

Some approximate formulas are presented for standard errors of estimated regression coefficients in two-level designs. If the researcher can make a reasonable guess as to parameters occurring in the model, this approximation can be a guide to the choice of sample sizes at either level. (SLD)

An equating design in which two groups of examinees from slightly different populations are administered a different test form with a subset of common items is widely used. This paper presents standard errors and a simulation that verifies the equation for large samples for an equipercentile equating procedure for this design. (Author/BS)

Regression to the mean (RTM) is often misunderstood. It is demonstrated that artifactual RTM depends fundamentally on the magnitude of measurement error at pretest. Adjustment usually involves estimating the measurement error and determining consequences, but even without adjustment, effects of RTM can be ameliorated. (SLD)

A generalized variance criterion is proposed to measure efficiency in item-response-theory (IRT) models. Heuristic arguments are given to formulate the efficiency of a design in terms of an asymptotic generalized variance criterion. Efficiencies of designs for one-, two-, and three-parameter models are compared. (SLD)

A methodology is presented for minimizing mean error variance in generalizability studies when resource constraints are imposed. The optimal number of observations and conditions of facets for random model, fully crossed one- and two-facet designs can be decided. Parallel closed form formulas can be determined for other designs. (SLD)

An analysis of covariance (ANCOVA) is done to correct for chance differences that occur when subjects are assigned randomly to treatment groups. When properly used, this correction results in adjustment of the group means for pre-existing differences caused by sampling error and reduction of the size of the error variance of the analysis. The…

The validity of various reliability-corrected procedures for adjusting for initial differences between groups in uncontrolled studies is established for subjects exhibiting linear fan-spread growth. The results are then extended to a nonlinear model of growth. (Author)

Textbook calculations of statistical power or sample size follow from formulas that assume that the variables under consideration are measured without error. However, in the real world of behavioral research, errors of measurement cannot be neglected. The determination of sample size is discussed, and an example illustrates blocking strategy.…

Standard errors and bias of unidimensional and multidimensional ability estimates were compared in a factorial, simulation design with two item response theory (IRT) approaches, two levels of test correlation (0.42 and 0.63), two sample sizes (500 and 1,000), and a hierarchical test content structure. Bias and standard errors of subtest scores…

Six models for evaluating change are examined via a Monte Carlo study. All six models show a lack of power. A modified analysis of variance procedure is suggested as an alternative. (JKS)

The framework and procedures of generalizability theory are introduced and illustrated in this instructional module that uses a hypothetical scenario involving writing proficiency. Generalizability analyses are useful for understanding the relative importance of various sources of error and for designing efficient measurement procedures. (SLD)