Identifying discriminating items for the student evaluation of teachers

Abstract

<p>The purpose of this study, conducted in 1986, was to develop and test student evaluation of teacher items suitable for completion of secondary and elementary school children. The items were designed utilizing current research on effective teaching behaviors. This study was conducted in two school districts and 12 school buildings in those districts. A total of 3,560 students participated in the study with data from 2,968 of those students being analyzed;Items for the questionnaire were developed after a review of the literature. The items were based on teacher behaviors that make a difference in student achievement and on valid, reliable, teacher evalution criteria. All items were tested for their reading level using two separate, computerized-readability programs;Four groupings of items, based upon readability, were developed. The four levels included items for completion by students in the lower elementary grades (grades K-2), the upper elementary grades (grades 3-6), the junior high (grades 7-8), and the high school (grades 9-12);Students in grades kindergarten through 12 completed questionnaires rating the presence and/or frequency of teacher behaviors in their classrooms using a five-point, Likert-type scale. The data from these questionnaires were analyzed using the Menne and Tolsma methodology to determine item discrimination power. Items which discriminated at the.05 level of significance were identified;The number of discriminating items range from 57 items at the high school level to 18 items for the upper elementary grades;The Cronbach alpha reliability coefficient was calculated to determine the internal consistency of all items with discrimination value equal to or exceeding 13 percent;All items on the four levels of the questionnaires that were found to be discriminating between teachers were then analyzed to determine the correlation between student responses on the discrimination questionnaires and four potentially biasing factors.</p>