Wayne W. Welch

A Test of a Model of Educational Productivity Among Senior High School Students.

Data from a national sample of 1,955 17-year-olds participating in the National Assessment in Science in 1981-82 were used to test a model of educational productivity involving ability, motivation,...

Curriculum Effects on the Social Climate of Learning: A New Representation of Discriminant Functions.

A primary goal of educational research has been to establish the conditions for effective learning in school classes. Much of such research has focused on classroom interaction in one form or another (See reviews by Medley and Mitzel, Remmers, and Withall and Lewis in Gage, 1963). We have probed this problem further by relating the perceived social climate to cognitive, affective, and behavioral learning (Anderson, 1968; Anderson and Walberg, 1968; Walberg and Anderson, 1968a). The present study, however, explores some potential determinants of the social climate itself in an effort to gain insight into the manner in which climate evolves. More specifically, the study investigates three questions in curriculum research: (1) What is the effect on the learning climate of a new physics course that provides for individual differences in learning? (2) What are the differences between the learning climates of classes of teachers with and without prior experience in teaching a new course? (3) How do the climates

A New Use of Randomization in Experimental Curriculum Evaluation

There has been no shortage of curriculum development in the United States during the past decade. In physics, mathematics, chemistry, and biology, for example, scientists and educators have worked together to design and develop many new conceptual frameworks, materials, and media. What has been given short shrift, however, is the evaluation of the new programs. There are a number of reasons: shortage of time, funds, and trained personnel; problems of experimenting with human subjects such as the Hawthorne effect; the problem of inference to populations from non-random samples; and so on. Solutions to some of these problems appear to be forthcoming, and some have already been proposed: more subtle and refined measurements, sophisticated statistical methods,2 strong inference,3 and efficient experimental designs.4 This paper is addressed to the latter solution, efficiency in experimental designs, as a consequence of random data collection from intact classrooms. Let us first consider the two traditional uses of randomization. As R. A. Fisher showed in 1925, the assumption underlying statistical inference is that the experiment to which it is applied meets the following conditions: (1) there has been a random selection of units from the population under study, from which population parameters can be estimated, and (2) for the estimation of experimental effects, there has been a random assignment of treatments to experimental units (and non-treatment to control groups). It would hardly seem necessary to point out these assumptions again in 1967, but edu