"Theory as Method: Small Theories of Treatments," by Mark W. Lipsey, is one of the most influential and highly cited articles to appear in "New Directions for Evaluation." It articulated an approach in which methods for studying causation depend, in large part, on what is known about the theory underlying the program. Lipsey discussed the benefits…

To predict suicide, explicit causal models of suicide outcome are needed in which predictors are temporally ordered and then tested statistically. Because suicide is a dichotomous, nominal scale outcome, only certain statistics for prediction are appropriate. Logistic regression, which utilizes the likelihood ratio test of statistical inference,…

Until experimenters can achieve some measure of control over access to information associated with their writing stimuli and outcome measures, research in composition will have a difficult time providing meaningful information on the unique effects on writing of such things as training programs, ethnicity, previous education, and sex. (HOD)

Examines difficulties with citation analysis as it is used to study citation frequency, usually for the evaluation of scientists, publications, or institutions. Topics addressed include selection of a control set of papers, comparisons of different types of papers (e.g., methodological or theoretical), effects of independent variables, and use of…

If they are to develop more effective social intervention strategies, social scientists must pay attention to social causality in their research. This will lead to more credibility for the social sciences and more support for research funds. Social causality is defined as a social process that produces a change in some dependent variable. It…

Discusses procedures and considerations involved with secondary analyses of longitudinal databases. Procedures involve (1) formulating questions; (2) creating a feasibility matrix; (3) reformulating questions; (4) creating derived variables; (5) performing data reduction; (6) analyzing data; and (7) interpreting results. Problems associated with…

The concept of quality is analyzed with special attention to its different senses in two contexts: qualitative evaluation and assessment of educational program quality. The view that evaluation of program quality cannot be validly expressed in quantitative terms and that qualitative evaluation methodologies have special relevance for the…

Explanations are offered for observed discrepant and null findings in the area of predictive information about teacher effectiveness. It is argued that if tautologies are to be avoided, only product variables are appropriate criteria of success. Some recommendations regarding future research are offered. (Author/GSK)

The authors state that multiple regression is a powerful method of statistical analysis, provides a strength of relationship index, and should replace analysis of variance (ANOVA) in educational research. They also discuss the coding of categorical variables and available computer programs for multiple regression. (Author/MH)

Reviews and critiques research that has assessed the effect of infant temperament on caregiver behaviors and concludes that the relationship between temperament and caregiving behavior has not been adequately tested. Offers conceptual and methodological guidelines for future research. (HOD)

Data in Yarworth and Gauthier's article on student self- concept and participation in school activities (EJ 189 606) were reanalyzed by Walsh and Winne (TM 505 375). Yarworth and Gauthier's criticism of the reanalysis (TM 505 376) is answered. (GDC)

Addresses the implicit assumption that instructional media represent meaningful experimental variables independent of instructional content or strategy. Results of computer-based instruction (CBI) studies (both well-designed and flawed) and meta-analyses are reviewed to support the authors' position that this assumption is invalid, and new…

Because of states' growing roles in education, the study of variation in state educational policies contributes to a better understanding of policy processes and to the prediction of possible policy outcomes. Comparing state educational policies involves choosing variables and analytical designs and defining cases, concepts, and measurement.…

Yarworth and Gauthier (EJ 189 606) examined whether self-concept variables enhanced predictions about students' participation in school activities, using unstructured stepwise regression techniques. A reanalysis of their data using hierarchial regression models tested their hypothesis more appropriately, and uncovered multicollinearity and…

This paper offers reasons for using traditional quantitative research methods and illustrates some of the considerations demanding rigor in a study on gender discrimination in school-administrator salaries. Traditional quantitative methods provide a meaningful language based on numbers and objectivity, and they are less time-consuming and complex…