Best linear prediction (BLP) and penalized best linear prediction (PBLP) are techniques for combining sources of information to produce task scores, section scores, and composite test scores. The report examines issues to consider in operational implementation of BLP and PBLP in testing programs administered by ETS [Educational Testing Service].

Propensity score analysis (PSA) is a methodological technique which may correct for selection bias in a quasi-experiment by modeling the selection process using observed covariates. Because logistic regression is well understood by researchers in a variety of fields and easy to implement in a number of popular software packages, it has…

With respect to the often-present covariance between error terms of correlated variables, D. W. Zimmerman and R. H. Williams's (1977) adjusted correction for attenuation estimates the strength of the pairwise correlation between true scores without assuming independence of error scores. This article focuses on the derivation and analysis of…

Under the assumption of normalcy, a formula is derived for the reliability of the maximum score. It is shown that the maximum score is more reliable than each of the single observations but less reliable than their composite score. (Author/LMO)

Three factors that increase score variability yet can be associated with an increase, a decrease, or no change in Pearson's correlation coefficient (r) are discussed (restriction of range, errors of measurement, and linear transformations of data). The connection between changes in variability and r depends on how changes occur. (SLD)

Reliability coefficients of linear combinations of observed scores have anomalous properties which have led to difficulties in the investigation of difference scores and gain scores in test theory. Discrepancies between classical results and correct results obtained from more general formulas, which allow for correlated errors, are examined…

Coefficient alpha, the most commonly used estimate of internal consistency, is often considered a lower bound estimate of reliability, though the extent of its underestimation is not typically known. Many researchers are unaware that coefficient alpha is based on the essentially tau-equivalent measurement model. It is the violation of the…

The objective of this paper is to examine the relationship between the unreliability of difference scores and the power of tests of significance in an attempt to determine the validity of the paradox for the measurement of change presented by Overall and Woodward: that the power of tests of significance is maximum when the reliability of the…

A method of estimating conditional standard error of measurement at specific score/ability levels is described that avoids theoretical problems identified for previous methods. The method focuses on variance of observed scores conditional on a fixed value of an observed parallel measurement, decomposing these variances into true and error parts.…

The degree of agreement between two raters rating several objects for a single characteristic can be expressed through an association coefficient, such as the Pearson product-moment correlation. How to select an appropriate association coefficient, and the desirable properties and uses of a class of such coefficients--the Euclidean…

Improvements are made on previous estimates for the conditional standard error of measurement in prediction, the conditional standard error of estimation (CSEE), and the conditional standard error of prediction (CSEP). Better estimates of how test length affects CSEE and CSEP are derived. (SLD)

Six methods for estimating the standard error of measurement (SEM) at specific score levels are compared by comparing score level SEM estimates from a single test administration to estimates from two test administrations, using Iowa Tests of Basic Skills data for 2,138 examinees. L. S. Feldt's method is preferred. (SLD)

This paper elaborates on the work of Cardinet, and others, by clarifying some points regarding calculations, specifically with reference to existing computer programs, and by presenting illustrative examples of the calculation and interpretation of several generalizability coefficients from a complex six-facet (factor) design. (Author/RL)

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)

Using data from a National Child Development Study (NCDS) in Great Britain, the applications of instrumental variable methods and structural equation models to estimating instrumental variables are presented. A subset of the longitudinal educational and home background data on children born in England, Wales and Scotland in a March week of 1958 is…