A general dichotomous condition is derived for the unidimensional polytomous Rasch model. The robustness of the dichotomous analysis is investigated in a simulation study. The model shows a close relation with the two-parameter Birnbaum model. (TJH)

A general latent variable model allows for maximum likelihood estimation with missing data. LISREL and LISCOMP programs may be used to carry out this estimation. Simulated data were generated. The proposed Full, Quasi-Likelihood estimator was found to be superior to listwise present quasi-likelihood and pairwise present approaches. (Author/GDC)

A complete enumeration method and a Monte Carlo method are presented to calculate the probability distribution of arbitrary statistics of adjacency matrices when these matrices have the uniform distribution conditional on given row and column sums, and possibly on a given set of structural zeros. (SLD)

A variety of paired comparison, triple comparison, and ranking experiments are discussed as generalized linear models. All such models can be easily fit by maximum likelihood using the GLIM computer package. Examples are presented for a variety of cases using GLIM. (SLD)

A Fisher's Z transformation is developed for the corrected correlation for conditions when the criterion data are missing because of selection on the predictor and when the criterion was missing at random, not because of selection. The two Z transformations were evaluated in a computer simulation and found accurate. (SLD)

Smallest exact confidence intervals for the ability parameter of the Rasch model are derived and compared to the traditional asymptotically valid intervals based on Fisher information. Tables of exact confidence intervals, termed Clopper-Pearson intervals, can be drawn up with a computer program developed by K. Klauer. (SLD)

This paper develops and applies three nonparametric comparisons of the shapes of two item characteristic surfaces: (1) proportional latent odds; (2) uniform relative difficulty; and (3) item sensitivity. A method is presented for comparing the relative shapes of two item characteristic curves in two examinee populations who were administered an…

Describes algorithms used in the computer program LOGIMO for obtaining maximum likelihood estimates of the parameters in loglinear models. These algorithms are also useful for the analysis of loglinear item-response theory models. Presents modified versions of the iterative proportional fitting and Newton-Raphson algorithms. Simulated data…

An adaptive testing algorithm is presented based on an alternative modeling framework, and its effectiveness is investigated in a simulation based on real data. The algorithm uses a latent class modeling framework in which assessed latent attributes are assumed to be categorical variables. (SLD)

Kernel smoothing methods for nonparametric item characteristic curve estimation are reviewed. A simulation with 500 examinees and real data from 3,000 records of the Graduate Record Examination illustrate the rapidity of kernel smoothing. Even when population curves are three-parameter logistic, simulation suggests no loss of efficiency. (SLD)

Asymptotic distributions of the estimators of communalities are derived for the maximum likelihood method in factor analysis. It is shown that equating the asymptotic standard error of the communality estimate to the unique variance estimate is not correct for the unstandardized case. Monte Carlo simulations illustrate the study. (SLD)

The algorithm developed by B. A. Schneider (1980) for analysis of paired comparisons of psychological intervals is replaced by one proposed by R. M. Johnson. Monte Carlo simulations of pairwise dissimilarities and pairwise conjoint effects show that Johnson's algorithm can provide good metric recovery. (SLD)

New methods for comparing the medians corresponding to independent treatment groups are suggested. Procedures are based on the estimator of F. Harrell and C. Davis in conjunction with a modification and extension of the bootstrap calibration technique suggested by W. Loh. Data from two groups provide an illustration. (SLD)

A class of models is presented for gamma distributed random variables. These additive, multiplicative, and combined additive-multiplicative models are more flexible than classical linear models with respect to the structure that can be imposed on expected values. As a special case, a class of psychometric models for reaction times is presented.…

A technique is presented to transform incomplete categorical data into complete data by imputing appropriate scores into missing cells. A solution of the optimization problem is suggested, and relevant psychometric theory is discussed. The average correlation should be at least 0.50 before the method becomes practical. (SLD)