NotesFAQContact Us
Search Tips
ERIC Number: ED356250
Record Type: Non-Journal
Publication Date: 1992-Oct
Pages: 55
Abstractor: N/A
Reference Count: N/A
The Effects of Data Truncation on Estimated Validity Indices for Course Placement. ACT Research Report Series 92-3.
Schiel, Jeff; Noble, Julie
Traditionally, correlation coefficients have been used to validate course placement decisions based on test scores and high school grades. Because placement systems restrict the range of both the predictor and outcome variables, correlation coefficients based on data from students enrolled in particular courses are understated relative to what they would be if placement had not occurred. Alternative methods therefore have been examined for validating placement systems. One such approach uses validity indices estimated from logistic regression analyses and distributions of predictor variables to determine placement effectiveness. The Assessment of Student Skills for Entry Transfer (ASSET) Basic Skills test scores and course grades of entering freshmen from 4 postsecondary institutions (49 students in accounting, 166 students in history at 2 schools, and 83 students in psychology) were analyzed to determine the impact of prior selection on the accuracy of estimated validity indices. Estimated validity indices based on truncated distributions of test scores and course grades were compared to the same indices based on full distributions. It was found that greater degrees of truncation are associated with a loss of accuracy in estimated validity indices. However, the loss of accuracy in the estimates was small when less than 15 percent of the data for the full distributions were truncated. Eight tables and 10 graphs illustrate the analysis. (Author/SLD)
ACT Research Report Series, P.O. Box 168, Iowa City, IA 52243.
Publication Type: Reports - Evaluative; Reports - Research
Education Level: N/A
Audience: N/A
Language: English
Sponsor: N/A
Authoring Institution: American Coll. Testing Program, Iowa City, IA.