Applied Psychological Measurement

Lord and Wingersky (1985) derived the asymptotic variance-covariance matrix for item and person param eters in item response models, using maximum likeli hood estimation. Their results can be used in incom plete designs, in which different test forms with common subtests are administered to different groups of examinees. It is also possible to estimate the accu racy of various designs beforehand, which enables the researcher to select the best of several designs under consideration. The possibilities are demonstrated for the one- and two-parameter models. Index terms: Asymptotic standard errors, Incomplete designs, Item banking, Information matrix, Maximum likelihood, Rasch model, Two-parameter logistic model.

This study investigated the similarity of information that is provided by direct and indirect methods of writing assessment. The skills required by each of these techniques provide a framework for a cognitive model of writing skills from which these procedures can be compared. It is suggested that practitioners in terested in reliably measuring all aspects of the pro posed writing process continuum, as characterized by this cognitive model, use both indirect and direct methods. Index terms: Confirmatory factor analy sis, Essay tests, Free-response tests, Multiple-choice tests, Writing assessment, Writing processes.

This paper has three purposes: (1) to develop a methodology based on axiomatic conjoint measure ment to accompany the McClelland (1980) fertility de cision-making model; (2) to demonstrate the use of the methodology by applying it to the study of a specific fertility-related behavior (contraceptive choice); and (3) to evaluate the usefulness and validity of the model and its methodology for studying fertility-re lated behaviors. The within-persons methodology de veloped is contrasted with the across-persons method ology customarily used to test expectancy-value models of behavior. Index terms: Conjoint mea surement, Contraception, Decision making, Expec tancy-value models, Preference models, Within-per sons analysis.

In restricted factor analysis each of the elements in the matrices (of factor loadings, factor correlations, and unique variances and covariances) can be fixed, free to be estimated, or constrained. In this paper it is argued that the common practice of constraining some parameters at 0 often is not psychologically meaning ful. Alternative procedures for constraining parameters are presented and illustrated using data based on the Luria model. Index terms: Constrained parame ters, Factor analysis, Factor correlations, Factor loadings, Factor uniqueness.

Hierarchical moderated regression (HMR) analysis may lead to interpretational problems in tests of mod erator (interaction) hypotheses. An alternative, resid ual-centering approach is described and compared to traditional HMR analysis. Procedures for evaluating in teraction hypotheses and general effect analysis proce dures are described for path (causal) models contain ing interactions. Index terms: Confirmatory analysis, Effect analysis in path models, Goodness-of- fit tests, Hierarchical moderated regression, Media tors, Moderated regression, Multicollinearity, Path models, Residual centering.

This study investigated the performance of five small-sample statistics (Lord's, Kristof's, McNemar's, Forsyth and Feldt's, and Braden's) that test whether two variables measure the same trait except for mea surement error. The conservative Type I error rates of the Lord and Kristof procedures and the liberal error rates of the McNemar, Forsyth and Feldt, and Braden procedures were corrected by determining appropriate critical values. Power comparisons were then made at the fixed α levels. In general, the McNemar statistic was shown to be the most powerful. Finally, the ef fects of non-normality were investigated, and it was demonstrated that the Braden technique became very liberal, whereas the other statistics tended to be some what liberal at the .01 significance level and reasona bly robust at the .05 level. Index terms: Disatten uated measures, Measurement error, Monte carlo simulation, Non-normality, Small-sample statistics, Type I error and power rates.

MOKSCAL is a program for the Mokken (1971) scale analysis based on a nonparametric item response model (Mokken & Lewis, 1982) that makes no assumption about the functional form of the item trace lines. The only constraint the Mokken model puts on the item trace lines is the assumption of double monotony; that is, the set of scalable items must have the properties of monotonicity in item difficulties and nondecreasing item trace lines implying a monotone relation between the item success probabilities and the latent trait (Sytsma & Molenaar, 1987). MOKSCAL tests this assumption.

In the present paper, linear programming was used to select items from item pools based on one-, two-, and three-parameter models so that a target test infor mation function was reached. The primary interest was in the distributional characteristics of the items thus selected. The results suggest that the linear program ming approach focuses on the "worst feature" of the target information function (i.e., the extremes of a uniform target and the maximum of a peaked target). The values of the parameters of the selected items tend to form clusters. For uniform targets, these clus ters are associated with the extremes of the target range, whereas for peaked targets they are associated with the maximum of the target. Selecting items from an item pool by linear programming appears to be a useful addition to the test constructor's repertoire. However, additional refinement may be needed to ob tain a specific distribution of item parameters for a given test. Index terms: Item response theory, Item selection, Linear programming, Target informa tion function.

Gulliksen's matched random subtests method is a graphical method to split a test into parallel test halves. The method has practical relevance because it maximizes coefficient α as a lower bound to the clas sical test reliability coefficient. In this paper the same problem is formulated as a zero-one programming problem, the advantage being that it can be solved by computer algorithms that already exist. It is shown how the procedure can be generalized to split tests of any length. The paper concludes with an empirical ex ample comparing Gulliksen's original hand-method with the zero-one programming version. Index terms: Classical test theory, Gulliksen's matched ran dom subtests method, Item matching, Linear program ming, Parallel tests, Test reliability, Zero-one pro gramming.