Michael C. Neale1
Randomized response is commonly used method to increase honest responding to sensitive questions on topics such as substance use or sexual abuse. For example, on every question the respondent tosses three coins in the air and if they are all heads they answer yes, otherwise they answer truthfully. Rates of positive responses have been shown to increase with this method, once corrected for the expected number of artificial positives. In the context of the classical twin study - or any study that focuses on the correlation between responses made in this fashion - some increase in the confidence interval of twin correlations is expected. This effect translates into reduced power to detect heritable and shared environmental variance. The loss of power is shown to be quite drastic for all except abnormally rare response mechanisms requiring a large number of coins. Inferences concerning the confidence intervals on heritability estimates for sensitive topics are drawn.
Address: Virginia Institute for Psychiatric and Behavioral Genetics, VCU, Box 980126, Richmond, VA 23298. Tel: 804 828 3369, Fax: 804 828 1471, Email: neale@psycho.psi.vcu.edu, Web URL: http://www.vipbg.vcu.edu/ ~neale/
1Virginia Institute for Psychiatric and Behavioral Genetics, VCU, Box 980126, Richmond, VA 23298. Supported by grants RR-08123 and MH-01458.