The study used earliest available student data from a flagship university in the southeast United States to build data mining models like logistic regression with different variable selection methods, decision trees, and neural networks to explore important student characteristics associated with retention leading to graduation. The decision tree…

The use of interaction and non-linear terms in multiple regression poses problems for determining parsimonious models. Several computer programs for using these terms are discussed. (JKS)

Decomposable models, which allow for the interdependence of structure among observable variables, are described. Each model is fully characterized by a set of conditional interdependence restrictions and can be visualized with an undirected as well as a special type of directed graph. (Author/JKS)

The Goodman technique of log linear analysis is ideal for family research, because it is designed for categorical (non-quantitative) variables. Variables are dichotomized (for example, married/divorced, childless/with children) or otherwise categorized (for example, level of permissiveness, life cycle stage). Contingency tables are then…

The purpose of the present study is to demonstrate the utility of data analysis methodology in evaluative research relating pupil and curriculum variables to pupil achievement. Regression models which account for achievement will result from the application of the methodology to two evaluative problems--one of curriculum comparison and another…

The purpose of this paper is to illuminate the advantages of path analysis for the exposition of results in data analytic papers. Probably the greatest advantage is that it provides a means by which the nature of the problem may be handily summarized. The method of path analysis, although conceived over sixty years ago by Sewell Wright, has only…

This paper, a revision of the original document, "Correcting Partial, Multiple, and Canonical Correlations for Attenuation" (see TM 000 535), presents the formula for correcting coefficients of partial correlation for attenuation due to errors of measurement. In addition, the correction for attenuation formulas for multiple and cannonical…

The basic objective of this research program was to develop and to test an accountability model of the educational process. In cooperation with the Pittsburgh School District, a major effort was devoted toward the development of a data base, the construction of appropriate models and the conduct of analyses of the data. Some analysis was…

Data accumulated over two years were analyzed in order to evaluate two components of a teacher education curriculum--performance objectives and self analysis skills. A multi-step procedure was devised and implemented which uses regression to illustrate the feasibility of implementing the procedure for curricular decision making. This procedure…

This final report summarizes past research and suggests new approaches to the problem of estimating long-term individual constants using path analysis. The general objective of the research was to detect and measure the likelihood that one variable, x, measured at time, t, has a causal influence on another variable, y, measured at a subsequent…

Path analysis, a technique related to multiple regression analysis is used for ascribing causal relationships among variables. Path analysis involves the construction of explicitly formulated causal models and makes the reasoning explicit in the form of path diagrams and structural equations. Regression analysis is then used to construct path…

Duncan's interpretation of his research on the nature and process of the distribution of rewards in the U.S. stratification system is questioned. Statistical and conceptual flaws are noted in Duncan's work and their implications explored. A reanalysis of primary survey data is presented and, based on a reinterpretation of this data, a…

Describes class exercises where undergraduates construct and test mathematical models using data and issues related to social security. The students collect data and forecast economic and demographic variables that determine future tax rates. They then discuss government policies that are derived from their models. (MJP)

The Joreskog and Sorbom LISREL (linear structural relations) method is investigated as an alternative to regression analysis in studies of aptitude-treatment interactions (ATI), to solve problems caused by unreliability of measurements and by large sets of variables. A study reported by M.J. Behr is reanalyzed. This study investigated relations…

This paper develops and estimates a model of the educational process. Most researchers have relied on a single equation, educational production function, which is an unrealistic formulation. A pupil's achievement, motivation, expectations, self-concept, and his perceived parents' and teachers' expectations are determined jointly, not…