Detection of quantitative trait loci associated with alcohol-dependence: Use of model-free sib-pair method and combined segregation-linkage analysis based on regressive models
Résumé
Two linkage methods were used to detect loci underlying neurophysiological measures associated with alcohol dependence 1) the Haseman-Elston (H-E) sib pair method for genome-wide search, and 2) the combined segregation-linkage (CSL), based on regressive models, to confirm positive linkages found by the genome screening. Among 14 linkage results that were significant at the 0.5% level using H-E, the CSL method leads to similar p-values in only three cases but to higher p-values in all others. Investigation of these discrepancies shows that assumptions (normality and homoscedasticity of the error term) of H-E least-squares regression method are not verified. A robust estimator of slope parameters without assuming any distribution function for the linear model error terms increases the p-values and reduces the difference between H-E and CSL results. Alternatively, the CSL approach may lack power when multiple genes with small effects are involved.