In IRT, item characteristic curves are created that represent different combinations of parameters such as item discrimination and item difficulty. In the logistic function of ICCs, it is assumed that the ICC comes to 0 on the probability axis.
I have made a little MINITAB macro that will allow one to enter the item discrimination and item difficulty values for 2 ICCs, and then see the two curves drawn (overlapping) to see what happens to the ICCs when parameters are changed. For students, running the macro a few times, varying a and b, one can quickly see what happens to the ICCs. NOTE: If you don't have Minitab, this will not work. Two examples can be seen at the following links: For difficulty comparisons: http://roberts.ed.psu.edu/users/droberts/diff.png For discrimination comparisons: http://roberts.ed.psu.edu/users/droberts/disc.png If you have a fairly new version of Minitab, you can also download the macro file ... that you can run in Minitab, to produce these kinds of curves. You would need to go to: http://roberts.ed.psu.edu/users/droberts/files/ and find the icc.mtb file. If you double click, you will see the actual file but, I think if you right click, you will get a dialog box or be able to SAVE this file ... in IE, save target as ... seems to allow this. (If you have a problem, you could highlight the actual text when the file opens up ... cut and paste to new file with the mtb extension) Once you have the file ... called icc.mtb ... you can open Minitab and, run this little EXEC macro. An example of what the actual macro looks like when you run it is: MTB > exec 'a:\icc' Executing from file: a:\icc.MTB This macro will let you compare 2 logistic ICCs at a time and will show you the 2 ICCs. For the first ICC, what is the discrimination value (a ... .3 to 2) and the difficulty value (b ... -3 to 3)? Enter each value on the next line, separated by a space. DATA> 1 0 MTB > Note: Now enter the discrimination value (a ....3 to 2) and MTB > Note: difficulty value (b ... -3 to 3) for the second ICC. DATA> 1 1 ------- I had entered a and b values of 1 and 0, and 1 and 1 ... for the 2 ICCs. When you do this, you will get nice plots of the type in the diff.png ... and disc.png ... links above. If anyone wants to try to use this file, and has a problem, I will be glad to try to help you. If you are using a MAC or something other than a standard Windows based program, you might be out of luck. Dennis Roberts, 208 Cedar Bldg., University Park PA 16802 <Emailto: [EMAIL PROTECTED]> WWW: http://roberts.ed.psu.edu/users/droberts/drober~1.htm AC 8148632401 . . ================================================================= Instructions for joining and leaving this list, remarks about the problem of INAPPROPRIATE MESSAGES, and archives are available at: . http://jse.stat.ncsu.edu/ . =================================================================
