Multiple regression analysis of sib-pair data on reading to detect quantitative trait loci

A simple extension of the DeFries and Fulker multiple regression model for twin analysis is applied to the problem of detecting linkage in a quantitative trait. The method, employing sib pairs, is based on that of Haseman and Elston. Reading data from 19 extended pedigrees were analyzed employing RLFPs as markers on chromosome 15 and using the widely available statistical applications software package, SAS. A number of possible linkages were detected, indicating that this approach is both powerful and effective, especially in the case of selected samples. Detecting genotype-environment interaction and the issue of power are briefly discussed. The programs used are available upon request.     

THE EFFECT OF SELECTED POOR ITEM-WRITING PRACTICES ON TEST DIFFICULTY, RELIABILITY AND VALIDITY

Violations of four selected principles of writing multiple choice items were introduced into an undergraduate political science examination. Three of the four poor practices had no overall effect on test difficulty. A significant (α= .05) interaction effect between the poor practices and course achievement occurred for one of the four practices, with the poorer students generally gaining most from the poorly written items. KR 20 values were significantly lower for sets of items with the same flaws than for "good" versions of the items in three of four comparisons. The reductions in reliability were equivalent to those expected to result from shortening the test by 13 to 56 percent. Concurrent validity (correlation of experimental test scores with final examination scores) was significantly lower in two of four cases. The reductions in validity were equivalent to those expected to result from shortening the test by 56 to 83 percent.