Classical methods for detecting outliers deal with continuous variables. These methods are not readily applicable to categorical data, such as incorrect/correct scores (0/1) and ordered rating scale scores (e.g., 0,..., 4) typical of multi-item tests and questionnaires. This study proposes two definitions of outlier scores suited for categorical…

There is significant potential for error in long production processes that consist of sequential stages, each of which is heavily dependent on the previous stage, such as the SER (Scoring, Equating, and Reporting) process. Quality control procedures are required in order to monitor this process and to reduce the number of mistakes to a minimum. In…

Researchers persist in using stepwise regression in spite of problems with this approach. As noted by B. Thompson (1995), three problems accompany the use of stepwise applications. The first is that computer packages may use incorrect degrees of freedom in their computations, resulting in a greater likelihood of obtaining a spurious statistical…

Some of the frustrations inherent in trying to incorporate qualifications of statistical results into meta-analysis are reviewed, and some solutions are proposed to prevent the loss of information in meta-analytic reports. The validity of a meta-analysis depends on several factors, including the: thoroughness of the literature search; selection of…

The first step in aggregating qualitative data is to establish parameters for cross site analysis. Establishing parameters should include: (1) building a conceptual framework, (2) defining units of analysis, (3) developing core questions, and (4) establishing guidelines for handling unique features in each site. Two basic approaches to analyzing…

This paper compares five growth measures using the High School and Beyond database. The measures are: (1) simple gain (posttest minus pretest); (2) difference between group means; (3) percentage of students who scored higher on the posttest than the pretest; (4) percentage of items missed on the pretest which are subsequently answered correctly on…

Three topics within the domain of regression-based analyses of multilevel data from quasi-experiments and field studies in educational research and evaluation are presented. The paper begins with a discussion of the general question of choice of unit of analysis, or what may be the more appropriate question of choice of analytical model. It then…

In almost all applications of measurement there is some sort of response by a human subject. Almost always, the response scale is ordinal, but almost always it is treated as if it were an interval measure. Methods for treating data ordinally are currently being developed in three areas: ordinal analysis for questionnaire responses, ordinal…

The major emphasis of the National Longitudinal Study of the High School Class of 1972 (NLS) has been directed toward individuals in the 1972 high school graduating class. Consequently, sampling parameters were established to ensure student representation, and the bulk of data collected were individual-level data. School-level data exist on a…

When random assignment has been accomplished and an analysis of covariance (ANCOVA) is being used to correct for initial differences among treatment groups, use of unreliable covariables not only decreases the power of ANCOVA, but also causes ANCOVA to test biased treatment effects. Several correction procedures have been suggested for the single…

The generation of questions of worthwhileness of academic advisement into researchable hypotheses is dependent upon results from outcome questions substantiating process questions. Researchable hypotheses are few due to the complexity of the advising process involving students, advisors, problems, methods, and researchers. Questions generated for…

Several procedures proposed in the literature for the analysis of growth curves are reviewed. Particular attention is given to the current issues in this area to guide practitioners in the selection of the most appropriate methodology. (Author)

Presents an investigation into the problem caused by using ordinal data in statistical procedures requiring at least interval-level data, and explores four elements of this problem. (MH)

Trade-offs are implicit in choosing a subjective or objective method for evaluating social programs. The differences between Bayesian and traditional statistics, decision and cost-benefit analysis, and anthropological and traditional case systems illustrate trade-offs in choosing methods because of limited resources. (SLD)

The adequacy of traditional effect size measures for research synthesis is challenged. Analogues to analysis of variance and multiple regression analysis for effect sizes are presented. The importance of tests for the consistency of effect sizes in interpreting results, and problems in obtaining well-specified models for meta-analysis are…