[Rasch] Setting up Winsteps to account for absent but valid scores
camelia.rosca at bc.edu
Thu Jun 24 09:51:04 EST 2010
I have a question regarding setting up Winsteps to take into account maximum scores even when they are not present in the data file.
The problem I have is as follows: I have a test for which some items are dichotomous, some are scored 0-2, some are scored 0-3, and some are scored 0-4. The problem is that most of the items scored 0-3 are so difficult (in the pre-test) that not even one student received a score of 3 (very few did in the post-test). So, when I run the pre-test data file through Winsteps the student measures are artificially inflated because I cannot get Winsteps to 'understand' that although the top score (3) is missing for some of the items, it should still consider it as a valid score. The CODES="01234" command does not help. I realized that Winsteps does not 'know' that there should be top scores of 3 because I made up data for a student and gave this imaginary student scores of 3 for the problematic items. After having done this the mean student measure dropped significantly (although the mean of this imaginary student was higher than the original mean student measure).
I tried to 'trick' Winsteps by recoding my numerical variables to strings in SPSS just so when I transfer the data to Winsteps I can rescore (NEWSCORE command) and specify the actual numerical values for each of the items but it seems that the fact that the actual value of 3 is missing takes precedence over specifying the possible values in the NEWSCORE command.
Has anyone encountered this problem with missing top scores in the past? How did you solve it?
Any advice or suggestion would be greatly appreciated.
Camelia Rosca, Ph.D.
Senior Research Associate
Center for the Study of Testing, Evaluation, and Educational Policy
camelia.rosca at bc.edu<mailto:camelia.rosca at bc.edu>
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the Rasch