Typically genetic studies of continuous traits such as cholesterol levels or blood pressure assume that interindividual variability follows a normal distribution. Here we develop methods to analyze positively skewed data by assuming a lognormal distribution. We develop a variance components approach for identifying such effects from a major gene, residual polygenic factors and nongenetic factors. We compare by a simulation study results from fitting this lognormal model with either applying the log transformation or not transforming the data. We found that the lognormal model provided more precise estimates and more powerful tests than a simple log transformation when analyzing lognormally distributed data. Power varied with sibship size. For the same total number of nonindependent sibpairs, larger sibships were less powerful. However, larger sibships are more economical because they require a smaller sample size to obtain a specified power. To illustrate the application of this lognormal model to real data, we studied evidence for linkage between triglycerides and the lipoprotein lipase gene.
ASJC Scopus subject areas