journal article
LitStream Collection
Home
Dunlap, William P.; Kennedy, Robert S.; Harbeson, Mary M.; Fowlkes, Jennifer E.
doi: 10.1177/014662168901300102pmid: N/A
This article demonstrates that slope and ratio scores may have the same psychometric difficulties—low re liability—as difference scores. Empirically, direct measures and derived scores from Baron's, Collins', Meyer's, and Posner's cognitive paradigms were ex amined in terms of their reliabilities and cross-correla tions. Reliabilities of the direct measures and their in tercorrelations were high. The derived measures, which were slope, ratio, and difference scores, had re liabilities near zero and, therefore, their cross-correla tions were also low. It is concluded that derived scores, although intuitively appealing as measures of mental operations, may have inherent psychometric difficulties that render them of little value for differen tial prediction. Index terms: cognitive paradigms, difference scores, individual differences, prediction, ratio scores, reliability, slope scores.
Schriesheim, Chester A.; Solomon, Esther; Kopelman, Richard E.
doi: 10.1177/014662168901300103pmid: N/A
LISREL maximum likelihood confirmatory factor analyses (Jöreskog & Sörbom, 1984) were conducted to explore the effects of two questionnaire formats (grouping versus randomizing items) on the conver gent and discriminant validity of two sets of question naire measures. The first set of measures consisted of satisfaction scales that had demonstrated acceptable psychometric properties in earlier studies; the second set of scales were job characteristics measures that had shown discriminant validity problems in previous re search. Correlational data were collected from two groups of employed business administration students (N = 80 in each group) concurrently (Study 1) and at two points in time (Study 2). The results of the anal yses showed that the grouped format was superior to the random format, particularly with respect to the weaker measures (the job characteristics scales). The results also illustrated and supported the usefulness of LISREL confirmatory factor analysis in studies of con vergent and discriminant validity. Index terms: con firmatory factor analysis, convergent validity, discrim inant validity, LISREL analysis, questionnaire formats, scale validity.
Woodruff, David J.; Sawyer, Richard L.
doi: 10.1177/014662168901300104pmid: N/A
Two methods are derived for estimating measures of pass-fail reliability. The methods require only a sin gle test administration and are computationally simple. Both are based on the Spearman-Brown formula for estimating stepped-up reliability. The non-distribu tional method requires only that the test be divisible into parallel half-tests; the normal method makes the additional assumption of normally distributed test scores. Bias for the two procedures is investigated by simulation. For nearly normal test score distributions, the normal method performed slightly better than the non-distributional method, but for moderately to se verely skewed or symmetric platykurtic test score dis tributions the non-distributional method was superior. Test results from a licensure examination are used to illustrate the methods. Index terms: Cohen's kappa, licensure examinations, pass-fail reliability, reliability, Spearman-Brown formula.
doi: 10.1177/014662168901300105pmid: N/A
When faced with the task of responding to a per sonality questionnaire, an individual may respond with a number of strategies or test-taking attitudes. Among these, deceptive (fake) and disengaged (random) atti tudes are of particular interest, for these can poten tially mislead and misinform test users. A two-stage model was devised to detect deceptive and disengaged protocols on the California Psychological Inventory. Using parameters from signal detection theory, this model is found to be highly sensitive in detecting in- validity. Index terms: California Psychological In ventory, expected utility, faking on personality inven tories, personality assessment, random response patterns, signal detection theory.
Mislevy, Robert J.; Stocking, Martha L.
doi: 10.1177/014662168901300106pmid: N/A
Since its release in 1976, Wingersky, Barton, and Lord's (1982) LOGIST has been the most widely used computer program for estimating the parameters of the three-parameter logistic item response model. An al ternative program, Mislevy and Bock's (1983) BILOG, has recently become available. This paper compares the approaches taken by the two programs and offers some guidelines for choosing between the two pro grams for particular applications. Index terms: Bayesian estimation, BILOG, IRT estimation procedures, LOGIST, marginal maximum likelihood, maximum like lihood, three-parameter logistic model estimation pro cedures.
doi: 10.1177/014662168901300108pmid: N/A
The accuracy of marginal maximum likelihood esti mates of the item parameters of the two-parameter lo gistic model was investigated. Estimates were obtained for four sample sizes and four test lengths; joint maxi mum likelihood estimates were also computed for the two longer test lengths. Each condition was replicated 10 times, which allowed evaluation of the accuracy of estimated item characteristic curves, item parameter estimates, and estimated standard errors of item pa rameter estimates for individual items. Items that are typical of a widely used job satisfaction scale and moderately easy tests had satisfactory marginal esti mates for all sample sizes and test lengths. Larger samples were required for items with extreme diffi culty or discrimination parameters. Marginal estima tion was substantially better than joint maximum like lihood estimation. Index terms: Fletcher-Powell algorithm, item parameter estimation, item response theory, joint maximum likelihood estimation, marginal maximum likelihood estimation, two-parameter logistic model.
Formann, Anton K.; Spiel, Christiane
doi: 10.1177/014662168901300109pmid: N/A
The linear logistic model with relaxed assumptions (LLRA) was developed for measuring changes in quali tative data. It assumes item-specific person parameters and thus does not require homogeneous items to be presented to the persons at two points in time. The hybrid variant of this model maintains the multidimen sionality of the person parameters, but it allows for different sets of items each of which is presented only once. In the model, a Rasch homogeneous item at t2 with possibly differing difficulty corresponds to each item at t1. A short description of both models is fol lowed by a first application of the hybrid LLRA to em pirical data from a study on text comprehension. This example not only serves to demonstrate possible re sults when applying the LLRA, but is also used to out line the principle of hypothesis testing and model con trols. Index terms: dichotomous data, linear logistic model, measuring change, Rasch model, text compre hension.
Showing 1 to 10 of 12 Articles