Stochastic approximation is suggested as a useful technique in areas where individuals have a goal firmly in mind, but lack sufficient knowledge to design an efficient, more traditional experiment. One potential area of application for stochastic approximation is that of formative evaluation. (CTM)

This paper outlines the problems associated with the estimation of variance components in generalizability analyses using analysis of variance software, and discusses the most useful software currently available for this specialist application: the MIVQUE method of the Statistical Analysis System (SAS) procedure VARCOMP. (Author)

The evaluation of competing analysis strategies based on estimator bias and variance is demonstrated using gains in standard scores and analysis of covariance procedures for quasi-experiments conforming to the fan-spread hypothesis. The findings do not lead to a uniform recommendation of either approach. (Author/JKS)

Five methods of performing pairwise multiple comparisons in repeated measures designs were investigated. Consideration of both Type I and Type II error rates found in the simulated conditions for the five procedures suggests that a Bonferroni method utilizing a separate error term for each comparison should be employed. (Author/JKS)

Computation and interpretation of Bayesian full-rank multivariate analysis of variance and covariance is described and illustrated in an exposition intended for readers familiar with univariate analysis of variance and multiple regression. (Author)

Browne's definitive but complex formulas for the cross-validational accuracy of an OSL-estimated regression equation in the random-effects sampling model are here reworked to achieve greater perspicuity and extended to include the fixed-effects sampling model. (Author)

The concerns of this investigation are multiple sources of complications in the analysis of multilevel educational data. Selected multilevel methods provide some indication of misspecification and can identify the direction of the bias in estimating teacher/class effects on mean class outcomes. (Author/CTM)

It is shown that using formulae for the estimation of sampling errors based on simple random sampling, when a design actually involves cluster sampling, can lead to serious underestimation of error. Jackknife and balanced repeated replication are recommended as techniques for dealing with this problem. (Author/CTM)

The effects of certain characteristics (e.g., sample design) of National Assessment of Educational Progress (NAEP) data on statistical analysis techniques are considered. Ignoring special features of NAEP data and proceeding with a standard analysis can produce inferences that underestimate the true variability and overestimate the true degrees of…