PDF pending restoration
ERIC Number: ED177173
Record Type: Non-Journal
Publication Date: 1978-May
Metric-Free Distributional Comparisons.
Haertel, Edward H.; And Others
Two methods are presented for comparing distributions, such as achievement test score distributions, for distinctly different groups of persons in such a way that the comparison will not be influenced by the particular metric of the test being used. Both methods use percentile scores. One method, attributed to Flanagan, fits a straight line to the graph representing the two distributions on normalized score scales. The second method finds the negative hypergeometric score distributions for each of the two groups that most closely approximate their observed score distributions. It has been shown that many observed test score distributions closely approximate the negative hypergeometric distribution with the same mean, variance, and number of items. These smoothed distributions are used as the basis of comparison between groups. An example is given in which the Flanagan line closely approximates the relationship shown by smoothed curves relating the distributions of two groups, where the median for one group was above the 93rd percentile of the other group. (CTM)
Publication Type: Reports - Research
Education Level: N/A
Authoring Institution: CEMREL, Inc., Chicago, IL. ML-GROUP for Policy Studies in Education.