Application of the graded response model to the assessment of student attitudes

This study demonstrates the application of the graded response model to the measurement of student attitudes towards mathematics. The purposes of the study were: (a) to investigate the application of Samejima's (1969) graded response model to attitude data, collected in the Third International Mathematics and Science Study (TIMSS) students' questionnaire, (b) to probe and investigate differences in male and female students' responses to an attitude scale using an IRT differential item functioning (DIF) procedure and, (c) to replicate the analyses in six jurisdictions that participated in TIMSS with the goal of exploring how these procedures work in different circumstances and to demonstrate the kinds of comparative results that can be obtained about the scales and about student responses to them. Student responses from six jurisdictions, US, Britain, France, New Zealand, Canada (French) and Canada (English), to the TIMSS student questionnaire provided the data for the application.; Factor analysis of polychoric correlations showed that two factors underlie the 19 items used in this study. Factor 1 comprised nine items related to students' "Reasons for doing well in mathematics" and Factor 2 comprised six items on which students expressed their "Personal perceptions of the nature of mathematics". The graded response model (Samejima, 1969) as implemented in PARSCALE 3.5 (Muraki & Bock, 1998) was used to calibrate the data, for each factor and for each jurisdiction separately. The results indicated that similar item parameters were obtained across the six jurisdictions. A review of the item parameters and item graphics showed that, whereas items on Factor 1 provided good measurements for students with low estimated theta values, the items on Factor 2 provided better measurements for those students with high estimated theta levels.; An important finding for mathematics educators is that two items on "Reasons for doing well in mathematics" factor favoured boys and one item on the "Personal perceptions of the nature of mathematics" factor favoured girls. A high correlation was also found between the traditional summated scores and IRT estimated theta scores.