A nonmetric algorithm, pairwise partitioning, is developed to identify feature-based similarity structures. Presents theorems about the validity of the features identified by the algorithm, and reports results of Monte Carlo simulations that estimate the probabilities of identifying valid features for different feature structures and amounts of…

The reliability of a gain is determined by the reliabilities of the components, the correlation between them, and their standard deviations. Reliability is not inherently low, but the components of gains in many investigations make low reliability likely and require caution in the use of gain scores. (SLD)

The critiques by L. Collins and L. Humphreys in this issue illustrate problems with the use of gain scores. Collins' examples show that familiar formulas for the reliability of differences do not reflect the precision of measures of change. Additional examples demonstrate flaws in the conventional approach to reliability. (SLD)

Explores the consequence of failing to incorporate measurement error in the development of cut scores in criterion-referenced measures, using the example of Texas and the Texas Assessment of Academic Skills to illustrate the impact of measurement error on false negative decisions. Findings support those of W. Haney (2000). (SLD)

Offers standard errors of prediction and confidence intervals for Vineland Adaptive Behavior Scales (VABS) that help in deciding whether variation in obtained scores of scale administered to the same person more than once is a result of measurement error or whether it reflects actual change in examinee's functional level. Presented values were…

Measurement of physical activity patterns is discussed in terms of data obtained by attitude assessment, activity questionnaires, personal monitoring devices, and fitness assessment. Problems of each technique are described. Application of activity measures to the estimation of total dietary needs is discussed. (SK)

When performing a classification study, it is often useful to leave a residue of unclassified entities to be analyzed separately. Using an interactional paradigm, theoretical reasoning for this approach is outlined. A procedure--RESIDAN--for conducting a classification analysis using a residue is described, and empirical data are provided. (TJH)

The sampling variability and zeta-values of parameter estimates for misspecified structural equation models were examined. A Monte Carlo study was used. Results are discussed in terms of asymptotic theory and the implications for the practice of structural equation models. (SLD)

Composite reliability and standard errors of measurement were computed for prorated Verbal, Performance, and Full-Scale intelligence quotient (IQ) scores from a seven-subtest short form of the Wechsler Adult Intelligence Scale-Revised. Results with 1,880 adults (standardization sample) indicate that this form is as reliable as the complete test.…

Comparison of responses to items pertaining to educational attainment on the Current Population Survey before and after 1992 found that the old (pre-1992) items led to more consistent measurement of precollege-level educational attainment. However, use of the old items leads to many errors in attributing degrees to individuals. (JOW)

Presents general principles for cleaning data sets. Suggests techniques: check for column shifts, data completeness, marginal frequencies, and selected cross-tabulations; examines a sample of cases. Gives strategies for preventing data processing errors. (SK)

Analyzed for examiner errors 55 Wide Range Achievement Test-Revised (WRAT-R) protocols completed by 9 practitioners for metropolitan school district. All practitioners made errors, which occurred on 95 percent of protocols and averaged 3.0 errors per protocol. Most frequent errors included failures to obtain correct ceiling or basal, and failures…

A numerical approach for computing standard errors (SEs) of a linear equating is described in which first partial derivatives of equating functions needed to compute SEs are derived numerically. Numerical and analytical approaches are compared using the Tucker equating method. SEs derived numerically are found indistinguishable from SEs derived…

In a previous issue, Lawrence Schweinhart and David Weikart conclude that children who attended a direct-instruction preschool program had significantly greater adult felony arrests than children following the High/Scope or nursery-school curricula. Analysis of research methods suggests improbable links between preschool experience and adult…

Studied whether moving average or autoregressive moving average models fit two longitudinal data sets previously thought to possess quasi-simplex structures better than the quasi-simplex, one-factor, or autoregressive models. Results of a Monte Carlo study show the importance of evaluating the fit, propriety, and parsimony of models before one…