Using three sets of real data, a comparison of four discrete discriminate analysis procedures is made using the actual versus the optimal error rate. The kernel method gives the most accurate results in all three cases. (Author/RL)

Suggests the importance of new paradigms in research and practice on testing English language learners (ELLs) to effectively address the complexities of language and culture, identifying three key issues: test review, test development, and treatment of language as a source of measurement error. Research examples illustrate the importance and…

A general longitudinal model is presented for estimating school effects and their stability. The model, capable of separating true changes from sampling and measurement error, controls statistically for effects of factors exogenous to the school system. The model is illustrated with data from large cohorts of students in Scotland. (SLD)

F. M. Lord's flexilevel, computerized adaptive testing (CAT) procedure was compared to an item-response theory-based CAT procedure that uses Bayesian ability estimation with various standard errors of estimates used for terminating the test. Ability estimates of flexilevel CATs were as accurate as were those of Bayesian CATs. (TJH)

Commonly employed measures of school quality can lead to very misleading results. Especially at the state level, nonrepresentative data such as aggregate Scholastic Aptitude Test scores provide very biased measures of school performance. Far superior are direct estimates of achievement growth. (SK)

Discusses problems involved with surveying and assessing literacy rates in developing nations. Defines current debates over literacy measurement, classification, and surveying. Examines and questions current assumptions linking literacy policy and economic development. Describes models for surveying and measuring literacy and numeracy skill…

Derivation of asymptotic variance-covariance matrices for item and personal parameters in item response models is demonstrated for one- and two-parameter models using maximum likelihood estimation. The results can be used in incomplete designs and in estimation of the accuracy of various designs beforehand. (TJH)

Results from this study indicate that a cutoff consisting of the mean plus a half standard deviation is more desirable than the original mean plus one standard deviation strategy for categorizing respondents into a "pure" identity status. Status-specific comparisons indicated groups were not significantly different on measures of…

Type I error rates of F. M. Lord's chi square test for differential item functioning were investigated using Monte Carlo simulations with marginal maximum likelihood estimation and marginal Bayesian estimation algorithms. Lord's chi square did not provide useful Type I error control for the three-parameter logistic model at these sample sizes.…

Techniques for adjusting census figures are discussed, with a focus on sampling error, uncertainty of estimates resulting from the luck of sample choice. Computer simulations illustrate the ways in which the smoothing algorithm may make adjustments less, rather than more, accurate. (SLD)

Addresses effects of reporting bias on spouses' answers to questions by comparing one-partner and couple data on marital violence. Results indicate significant underreporting of violence, and that some predictors of violence are significantly related to underreporting by one spouse. (JPS)

This article cautions against complacency in "subjective" assessment, arguing that even tests designed to reflect the development of learner-centered, interactive and communicative approaches to teaching English may have cultural bias built into their assessment criteria. The reply article singles out as an unresolved issue whether or…

In a meta-analysis of the effect of adding hypnosis to cognitive-behavioral treatments for weight reduction, additional data were obtained from authors of two previous studies, and computational inaccuracies in the previous meta-analyses were corrected. Discusses findings. Correlational analyses indicated that the benefits of hypnosis increased…

The basic theory of Rasch measurement applies to situations where a person has a certain level of a trait being investigated, and this level of ability is what determines (to within a measurement error) how well the person does on each item in a test. This paper responds to frequent suggestions from colleagues that the use of Rasch measurement…

Extensive evidence exists demonstrating the utility of Curriculum-Based Measurement in reading (R-CBM) for progress-monitoring purposes; however, most studies have evaluated R-CBM from a traditional psychometric perspective, which allows for variability in individual student's data that is not a function of increased skills (i.e., measurement…