How can maximum likelihood methods reveal candidate gene effects on a quantitative trait?