Presents a preliminary empirical test of a maximum likelihood approach to using relevance data for training information retrieval parameters. Discusses similarities to language models; the unification of document-oriented and query-oriented views; tests on data sets; algorithms and scalability; and the effectiveness of maximum likelihood…

In the approach to latent trait models for pure speed tests presented in this article, the subject parameters are treated as random variables with a common gamma distribution, and marginal maximum likelihood estimators are derived for the test difficulties and the parameters of the latent subject distribution. An application of this model to…

Misspecification of mean and covariance structures for metric endogenous variables is considered. Maximum likelihood estimation of model parameters and the asymptotic covariance matrix of the estimates are discussed. A Haussman test for misspecification is developed, which is sensitive to misspecification not detected by the test statistics…

Various aspects of the computerized adaptive testing (CAT) procedure for partial credit scoring were manipulated, focusing on the effects of the manipulations on operational characteristics of the CAT. The effects of item-pool size, item-pool information, and stepsizes used along the trait continuum were assessed. (TJH)

A sequential-regimes job search model tested the effect of advanced notice on the duration of joblessness. Maximum likelihood estimates using data from the 1984 and 1986 Dislocated Worker Surveys demonstrated that advance knowledge significantly shortened joblessness for most labor force groups. (SK)

A simulation study by B. D. Wright and G. A. Douglas is critiqued, which indicates that the unconditional maximum likelihood method is an appropriate substitute for the theoretically superior conditional method for estimating parameters of the Rasch model. The study appears to rest on inadequate logic. (TJH)

L. A. Goodman's loglinear formulation for bi-way contingency tables is extended to tables with or without missing cells and is used for exploratory purposes. Three-way tables and generalizations of correspondence analysis are deduced, and a generalized version of Goodman's algorithm is used to estimate scores in all cases. (Author/TJH)

The nonnormed fit index's dependence on sample size in covariance structure analysis is discussed. Contrary to K. A. Bollen (1986) (whose alternative index depends on sample size), it is shown that the mean of the nonnormed fit index is independent of sample size for true and almost true models. (Author/TJH)

Using a 21-statement attitude-type instrument, an iterative procedure for improving confirmatory model fit is demonstrated within the context of the EQS program of P. M. Bentler and maximum likelihood factor analysis. Each iteration systematically eliminates the poorest fitting statement as identified by a variable fit index. (SLD)

A Monte Carlo experiment was conducted to investigate the performance of bootstrap methods in normal theory maximum likelihood factor analysis when the distributional assumption was satisfied or unsatisfied. Problems arising with the use of bootstrap methods are highlighted. (SLD)

Analysis of the pairwise preferences of pairs of gambling stimuli elicited from each of 40 chronic schizophrenic inpatients, 32 schizophrenic outpatients, and 33 siblings or parents of these patients indicates that choice behavior was usually determined by the expected value of the stimuli, with no significant differences among groups. (SLD)

Introduces and discusses the rationale and procedures of two nonparametric approaches to estimating the operating characteristic of a discrete item response, or the conditional probability, given the latent trait, that the examinee's response be that specific response. (SLD)

Discusses fitting multivariate normal mixture distributions to structural equation modeling. The model used is a LISREL submodel that includes confirmatory factor and structural equation models. Two approaches to maximum likelihood estimation are used. A simulation study compares confidence intervals based on the observed information and…

Generalizing Boolean matrix decomposition to a larger class of matrix decomposition models is demonstrated, and probability matrix decomposition (PMD) models are introduced as a probabilistic version of the larger class. An algorithm is presented for the computation of maximum likelihood and maximum a posteriori estimates of the parameters of PMD…

The Rasch Poisson counts model is a latent trait model for the situation in which "K" tests are administered to "N" examinees and the test score is a count (repeated number of some event). A mixed model is presented that applies the EM algorithm and that can allow for missing data. (SLD)