To the Editor.
—We read with great interest the articles by Bohman, Cloninger, and Sigvardson entitled "Inheritance of Alcohol Abuse" and "Maternal Inheritance of Alcohol Abuse" (1981;38:861-868 and 1981;38:965-969). The authors have been fortunate in their access to a large data set on adoptees, which gives them enormous potential to answer some important questions regarding the issue of nature v nurture in the genesis of alcoholism.However, we are concerned that the authors have not used the best possible statistical methods. There appear to be important shortcomings in their data analysis.Given that the authors are interested in discovering which factors affect the probability of alcohol abuse, it would have been more appropriate to use logistic regression with the genetic and environmental variables as the "independent" variables. This is a direct way of assessing which variables are important. Additionally, this type of model can be used to estimate the effects