Two Experiments in Learning Educational Statistics

Abstract

Do students obtain most from courses in educational statistics by spending all of their time in working problems, in reading statistical textbooks and literature, or in a combination of both of these procedures ! In order to secure information on this question the writer conducted his course in educational statistics last year in an experimental manner. The eighteen students in the class, all of whom were graduate students, were divided into three groups having equal average intelligence as measured on two group intelligence tests. Upon entering the class only three of the students had any knowledge of educational statistics. One of these stu dents was placed in each of the three groups. Each student in the reading group was to read two hours daily from a carefully prepared reading list, and prepare a report showing the time and the topics covered. Each student in the problems group was to spend two hours each day working statistical problems. This work was handed in each day. Each member in the combination group was to spend one hour daily reading and one hour in working problems. A check on the amount of time spent on the assignments indicated that each group spent an average of two hours and two minutes daily. The experiment wyas conducted according to this plan for six weeks during which time three true-false tests were given. At the end of this period two objective-type achievement tests in educational statistics (pre pared by the author) were given, namely, Ri (a test of interpretations) and Pi (a test of processes.) For the second six-weeks of the experiment the reading and problems groups were interchanged, the combination group remaining the same. During this period two true-false tests were given, answers to which were to be given both in and out of class (giving two sets of results). This made a total of seven true-false tests during the experiment. At the end of the second six-weeks period tests Ra and Pu (separate forms of the test on interpretations and the test on processes, respectively) were given. The authors of the two intelligence tests used to obtain equivalent groups claim a rather substantial reliability for the respective tests. Fifteen of the twenty-one intercorrelations computed for all of the possible combinations of the scores on the seven true-false tests, are sta 674