It's not one or the other. I look harder for possible ambiguities in a question if the difficulty & discrimination indices are out of whack. FWIW, I try to emphasize to the class that my adjustments are not a "curve." Instead, I am introducing a "correction factor" for bad questions. I suspect that the difference is lost on them but it makes me feel better. And to give Michael an "amen": both he and I have reputations as being pretty tough and neither of us has a problem with grade inflation. That makes it a lot easier to play "Mr. Niceguy." Ed ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ Edward I. Pollak, Ph.D., Department of Psychology, Husband, father, grandfather, biopsychologist, herpetoculturist and bluegrass fiddler.... not necessarily in order of importance. ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ > Date: Sun, 8 Apr 2001 12:37:40 -0500 > From: Chuck Huff <[EMAIL PROTECTED]> > Subject: MC thanks & discrimination index > Message-ID: <v04220803b6f650e0a5f8@[209.163.34.25]> > > Thanks to both Ed Pollack and Michael Renner for suggesting the > following criteria: > > 1) A cutoff of .25 or .33 of those who got it correct > 2) The item to whole correlation as a discrimination index > 3) Asking for what students were thinking of when they got it wrong. > > I like 3 the most, since it helps me figure out why the question may > have been misleading. > > I am still not sure how to interpret the item-to-whole correlation, > though. Is it bad only if it is negative?
