In a LISREL-IV analysis, a method of specifying a priori the variances of the latent variables for interpretability is demonstrated. The potential confusion of the metric of the latent variables is discussed, since many of the parameter estimates are a function of the metric. (Author/CM)

Simple modifications of principal component methods are described that have distinct advantages for structural analysis of relations among educational and psychological variables. The methods are contrasted theoretically and empirically with conventional principal component methods and with maximum likelihood factor analysis. (Author/GK)

Reviews methods for handling missing data in a research study. Model-based methods, such as maximum likelihood using the EM algorithm and multiple imputation, hold more promise than ad hoc methods. Although model-based methods require more specialized computer programs and assumptions about the nature of missing data, these methods are appropriate…

A necessary and sufficient condition is given in this paper for the existence and uniqueness of the maximum likelihood (the so-called joint maximum likelihood) estimate of the parameters of the Partial Credit Model. This condition is stated in terms of a structural property of the pattern of the data matrix that can be easily verified on the basis…

The Akaike Information Criterion (AIC) was introduced to extend the method of maximum likelihood to the multimodel situation. Use of the AIC in factor analysis is interesting when it is viewed as the choice of a Bayesian model; thus, wider applications of AIC are possible. (Author/GDC)

Typically, nonlinear models such as those used in the analysis of covariance structures, are not globally identifiable. Investigations of local identifiability must either yield a mapping onto the entire parameter space, or be confined to points of special interest such as the maximum likelihood point. (Author/JKS)

Studied whether the standard z-statistic that evaluates whether a factor loading is statistically necessary is correctly applied in such situations and more generally when the variables being analyzed are arbitrarily rescaled. An example illustrates that neither the factor loading estimates nor the standard error estimates possess scale…

A Bayesian procedure is given for estimation in unrestricted common factor analysis. A choice of the form of the prior distribution is justified. The procedure achieves its objective of avoiding inadmissible estimates of unique variances, and is reasonably insensitive to certain variations in the shape of the prior distribution. (Author/BJG)

Modeling the population in survey sampling problems continues to be controversial. An important reason is that the likelihood principle makes it somewhat necessary to model the population. Estimating the population total in two-stage survey sampling is considered, making use of a "superpopulation" model. The problem is then really one of…

An exploratory study of the value of ridge regression for interactive models is reported. Assuming that the linear terms in a simple interactive model are centered to eliminate non-essential multicollinearity, a variety of common models, representing both ordinal and disordinal interactions, are shown to have "orientations" that are…

When the results of a series of independent studies are combined, it is useful to quantitatively estimate the magnitude of the effects. Several methods for estimating effect size are compared in this paper. Glass' estimator and the uniformly minimum variance unbiased estimator are based on the ratio of the sample mean difference and the pooled…

Estimation of ability and item parameters in latent trait models is discussed. When both ability and item parameters are considered fixed but unknown, the method of maximum likelihood for the logistic or probit models is well known. Discussed are techniques for estimating ability and item parameters when the ability parameters or item parameters…

Equivalence of marginal likelihood of the two-parameter normal ogive model in item response theory and factor analysis of dichotomized variables was formally proved. Ordered and unordered categorical data and paired comparisons data were discussed, and a taxonomy of data for the models was suggested. (Author/GDC)

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)

Known results for testing and estimation problems for patterned means and covariance matrices with explicit linear maximum likelihood estimates are applied to the block compound symmetry problem. An example involving educational testing is provided. (Author/JKS)