Use of the Mantel-Haenszel procedure as a test for differential item functioning under the Rasch model of item-response theory is examined. Results of the procedure cannot be generalized to the class of items for which item-response functions are monotonic and local independence holds. (TJH)

Recently, C. E. McCulloch (1987) suggested a modification of the Morgan-Pitman test for comparing the variances of two dependent groups. This paper demonstrates that there are situations where the procedure is not robust. A subsample approach, similar to the Box-Scheffe test, and the Sandvik-Olsson procedure are also assessed. (TJH)

Aggregate scoring was applied to a recertifying examination for medical professionals to generate an answer key and allow comparison of peer examinees. Results for 1,927 candidates for recertification indicate considerable agreement between the traditional answer key and the aggregate answer key. (TJH)

A solution is offered to problems associated with the inequality in the manipulability of probabilities of classification errors of masters versus nonmasters, based on competency test results. Eschewing the typical arbitrary establishment of observed-score standards below 100 percent, the solution incorporates a self-correction of wrong answers.…

For Y. Yao's and G. S. James' tests, Type I error rates were estimated for various combinations of the number of variables, sample-size and sample-size-to-variables ratios, and heteroscedasticity. These tests are alternatives to Hotelling's T(sup 2) and are intended for use when variance-covariance matrices are unequal for two independent samples.…

Compares results of Hunter-Schmidt meta-analytic technique with results of Glass meta-analytic technique on three meta-analytic data sets chosen from the literature, hypothesizing that the Hunter-Schmidt mean effect size would be significantly larger than the Glass mean effect size because of correlation for measurement error. Results confirmed…

Simplified formulas are proposed for computing the standard errors of equipercentile equating for continuous and discrete test scores. These formulas are easily extended to more complicated equating designs. Results from a study of 719 subjects taking an English test indicated that the formulas work reasonably well for moderate-size samples. (SLD)

Four general criteria are proposed for the choice of a metrical solution for a causal effect: (1) compatibility with the effect; (2) ease of communication; (3) lack of measurement error bias; and (4) stability across subjects and situations. These criteria are illustrated for randomized and nonrandomized designs. (SLD)

The distinction between two models of reliability is clarified. Reliability may be conceived of and estimated from a true score model or from the perspective of sampling precision. Basic models are developed and illustrated for each approach using data from the author's work on measuring organizational climate. (SLD)

Some of the threats to internal validity that arise when evaluating the impact of water conservation programs during a drought are illustrated. These include differential response to the drought, self-selection bias, and measurement error. How to deal with these problems when high-quality disaggregate data are available is discussed. (SLD)

Effects of response guided experimentation in applied behavior analysis on Type I error rates are explored. Data from T. A. Matyas and K. M. Greenwood (1990) suggest that, when visual inspection is combined with response guided experimentation, Type I error rates can be as high as 25%. (SLD)

The study calculated indices of interjudge reliability and interjudge and intrajudge agreement on ratings made by 15 experienced speech clinicians on 5 deviant speech dimensions of 15 speakers with ataxic dysarthria and a wide range of speech intelligibility. Judges were reliable in tracking imprecise consonants, excess and equal stress, and harsh…

Nonlinear relationships and latent variable assumptions can lead to serious specification errors in structural models. A quadratic relationship, described by a linear structural model with a latent variable, is shown to have less predictive validity than a simple manifest variable regression model. Advocates the use of simpler preliminary…

A reparameterization is formulated that yields estimates of scale-invariant parameters in recursive path models with latent variables, and (asymptotically) correct standard errors, without the use of constrained optimization. The method is based on the logical structure of the reticular action model. (Author)

The relationship between self- and peer report on the Coping Inventory for Stressful Situations (CISS) (N. Endler and J. Parker, 1990) was studied with 163 pairs of friends (college students). Positive but modest correlations were found between peer and self-report for three types of coping. Findings that attenuate these correlations are…