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…

Selectivity problems are discussed in terms of a general model that is estimated by the maximum likelihood method. Both single-group and multiple-group analyses are considered. An extension of the general model to latent variable models is discussed. (Author/PN)

Recently, C. E. McCulloch (1987) suggested a modification of the Morgan-Pitman test for comparing the variances of two dependent groups. This paper demonstrates that there are situations where the procedure is not robust. A subsample approach, similar to the Box-Scheffe test, and the Sandvik-Olsson procedure are also assessed. (TJH)

Examined the performance of a recently available full information maximum likelihood (FIML) estimator in a multiple regression model with missing data using Monte Carlo simulation and considering the effects of four independent variables. Results indicate that FIML estimation was superior to that of three ad hoc techniques, with less bias and less…

This study of maximum likelihood confirmatory factor analysis found effects of practical significance due to sample size, the number of indicators per factor, and the number of factors for Joreskog and Sorbom's (1981) goodness-of-fit index (GFI), GFI adjusted for degrees of freedom, and the root mean square residual. (Author/BW)

The application of latent trait theory to classroom tests necessitates the use of small sample sizes for parameter estimation. Computer generated data were used to assess the accuracy of estimation of the slope and location parameters in the two parameter logistic model with fixed abilities and varying small sample sizes. The maximum likelihood…

Single sample approximations are considered for the cross-validation coefficient in the analysis of covariance structures. Results of a random sampling experiment--using data from ability tests administered to high school students (sample sizes 100, 400, and 800)--illustrate the coefficient and adjustment for predictive validity. (SLD)

Estimating item parameters from a two-parameter normal ogive model is considered using Gibbs sampling to simulate draws from the joint posterior distribution of ability and item parameters. The method gives marginal posterior density estimates for any parameter of interest, as illustrated using data from a 33-item mathematics placement…

A method is proposed for constructing a population covariance matrix as the sum of a particular model plus a nonstochastic residual matrix, with the stipulation that the model holds with a prespecified lack of fit. The procedure is considered promising for Monte Carlo studies. (SLD)

Data are summarized in Scheuneman's Score x Group x Response frequency table in order to investigate item bias. The data can arise from two different sampling models: (1) multinomial sampling in which a fixed sample size is used and the responses are cross-classified according to score, group, and response; and (2) product-multinomial sampling in…

Several models for optimizing incomplete sample designs with respect to information on the item parameters are presented. The following cases are considered: (1) known ability parameters; (2) unknown ability parameters; (3) item sets with multiple ability scales; and (4) response models with multiple item parameters. The models are able to cope…

A three-level multivariate statistical modeling strategy is presented that resolves the question of whether the unit of analysis should be the teacher or the student. A reanalysis of U.S. high school data (51 Catholic and 59 public schools from the High School and Beyond survey) illustrates the model. (SLD)