Effects of censored variables on family studies

In the present paper effects of censored variables on estimates of genetic and environmental influences were studied. Analytic results showed with 50% censoring, about 15-20% of the variance may be attributed to the wrong source and that this amount increases rapidly with more than 70% censoring. Censoring effects on comparisons between different genetic studies, subgroups within a study (e.g., sex or age groups), or different behaviors (e.g., the heritabilities of delinquency and depression) were also examined. Results indicated that censoring may be quite influential for these kinds of comparisons. For instance, it was demonstrated that, especially for unstandardized solutions, small initial group differences in means can lead to seriously biased conclusions concerning the resemblance in biometric parameters. Finally, a stimulation study supported the applicability of the general analytical results and showed that summed scores of censored Likert-type items may be seriously affected by censoring.