Regional inference with averaged P values increases the power to detect linkage.

Controversy exists with respect to the choice of an appropriate critical value when testing for linkage in a genomic screen. A number of critical values have been proposed for single-locus and multi-locus linkage analyses. In this study, criteria based on multiple single-locus analyses (i.e., regional test criteria) are evaluated using simulation methods for three different map densities. Tests based on single loci, multiple consecutive single loci, and moving averages of consecutive single loci are considered. Appropriate critical values are determined based on results from simulations under the null hypothesis of no linkage. The power of each "regional test " was compared to the power of a single-locus test. Results suggest that the best power was found when averaging P values over an interval size of 9-15 cM, and that testing the average of P values from two consecutive loci is superior to testing each single locus separately. The increase in power ranged from 7- 29% over the simulations considered. Copyright 1999 Wiley-Liss, Inc.