A simple method for censored age-of-onset data subject to recall bias: mothers' reports of age of puberty in male twins.

Genetic analysis of variation in age of onset of development milestones or psychopathological behaviors has been little researched, owing largely to the computational difficulty of dealing with "censored" observations. Censored observations arise when the only information on individuals is that they have reached a particular age but without onset having occurred. This paper shows how models can be simply fitted to such data using programs that can perform genetic analysis of categorical data by maximum likelihood. The method is illustrated using the program Mx with data on maternal report of the onset of puberty in twin sons from the Virginia Twin Study of Adolescent Behavioral Development. Frequently, data on age of onset is collected by retrospective recall. This can pose a variety of measurement problems. Suggestions are made for models that account for some of these problems or are robust to their presence. Substantial evidence for "telescoping" of onset dates is found for the puberty data. If left unaccounted for, these effects can artifactually inflate estimates of common environment effects.