Re: Reading Ability and Grades

[Robert Herdegen]

At 09:00 AM 3/25/2002 -0800, you wrote:
>Recently, I signed on to do some research for my college. At present I am
>analyzing existing data and will move on to a more thorough examination of
>various factors. My first findings were interesting enough that I thought to
>pass them along to TIPS.
>Sierra uses the Nelson Denny Reading Test as an assessment tool. The data I
>was given was for Reading Assessment Scores from the Fall of 1997 and the
>course performance through 1999. The scores were compared to the performance
>on approximately 6,000 courses. The correlation beween assessment scores and
>performance was around +.85 which is impressive. However, consider this: the
>range of scores on the assessment was from 2-35. Students scoring from 30-35
>completed (grade of C or better) 77% of their courses. but students scoring
>from 5-10 passed about 50% of their classes. I would maintain that using
>this assessment test as a screening for students is inappropriate. My idea
>is to give the counselors a tool. I did a simple linear regression and got
>95% confidence interval estimates of the range of completion for each score.
>I would like to provide these to our counselors who could then inform
>students who took the test what their score meant in terms of likelhood of
>passing the class. A student who got a 10, for example, would have a 50%
>chance of passing with a range of 25 to 75%. The student could then decide
>whether to take the class. Comments?
>
>By the way the courses sample were history, psychology, business and
>philosophy.
>
>
>
>
>Harry Avis PhD
>Sierra College
>Rocklin, CA 95677
>email: [EMAIL PROTECTED]

Harry:

This may not be exactly what you are looking for, but is something to think about. One thing you may want to do is to perform a curvilinear regression on the data. Recently I did some work looking at a variety of predictors of college GPA, including SAT scores, high school GPA, weighted high school GPA, etc., etc. (many of the usual suspects, and a few new ones). A curvilinear regression using a simple polynomial function produced a much higher R-square than a simple linear regression for each of the key variables. In each case the regression line was a positively accelerated function, virtually flat for lower half of the range of scores on the predictor variable, and positively sloped for the upper half of the range. The results are, perhaps, as expected. The predictors of academic performance don't seem to predict very well for the weaker students, where perhaps things like motivation and "cognitive style" may be more critical determinants of performance than aptitude. Thus, they are not particularly useful in predicting performance for the group for whom prediction of performance is most important. The point? If the Nelson-Denny performs like these other factors, using it as an assessment tool for identifying at-risk students or for predicting how students will perform in a particular domain may be inappropriate, because it does not predict for the weakest students.

Bob
****************************************************************
Robert T. Herdegen III
Elliott Professor of Psychology
Department of Psychology
Hampden-Sydney College
Hampden-Sydney, VA 23943
434-223-6166
****************************************************************
--- You are currently subscribed to tips as: [EMAIL PROTECTED] To unsubscribe send a blank email to [EMAIL PROTECTED]