Application of Rasch measurement theory to the Torrance tests of creative thinking, figural form a
MetadataShow full item record
Five overarching assumptions are currently made about the items in the Torrance Tests of Creative Thinking, Figural Form A (TTCT-figural Form A; Torrance, 1966). The items are assumed to have a good fit, to be equally difficult, to be composed of equally distributed response options, to behave the same for each gender, and to involve the same amount of standard error of measurement which is at the same level along the latent trait continuum. In the present study, Rasch measurement theory was utilized to examine the items in the TTCT-figural Form A for each of these assumptions regarding all the variables measured in the test, as well as item-level elaboration (i.e., elaboration-I, a variable fabricated for the present study). Data were collected from 193 second grade students in Turkey. The dichotomous Rasch model and the rating scale model were used for analyses. It was found that all the items had a good fit regarding 14 variables. Misfit was detected for three items in total regarding three variables. The items were not estimated to be at the same difficulty level regarding any of the variables except elaboration-I and humor. The response options of the items (or activities) were distributed evenly regarding resistance to premature closure but not elaboration, elaboration-I, and abstractness of titles. The items were estimated to behave the same for each gender regarding the majority of the variables. Differential item functioning was detected for eight variables. Finally, it was estimated that all the items involved the same amount of standard error of measurement and that the amount of standard error of measurement involved in an item increased towards the far ends of the latent trait continuum. Overall, the findings suggested that the items in the TTCT-figural Form A possessed sufficient quality for providing appropriate person measures for both genders regarding the majority of the variables. The item reliability indexes for 14 variables indicated a high possibility that invariant item calibration was attained regarding those 14 variables. Invariant measurement of the trait levels of the students, on the other hand, was not achieved regarding any of the variables.