9
$\begingroup$

A statistician friend of mine told me of an interesting technique used to obtain honest responses on surveys that dealt with sensitive issues. I recall the general gist of the method, but am wondering if anyone knows the details and if it is referenced anywhere.

The story was that the Florida AMA wanted to assess drug use among physicians. They sent out a questionnaire with one die. IIRC, the instructions were something along the lines of "roll the die. If you have ever taken drugs or get a 6, write down six; otherwise write down whatever number comes up". The idea being that if someone pulled the doctor's questionnaire and saw a six, he could say he didn't take drugs, he just happened to roll a six.

$\endgroup$
  • $\begingroup$ this very question hints at the "personal is political" motto of some years back, where everyone was stoned and everything was political, now everyone can still be stoned, but we try to downplay the political repercursions ;) $\endgroup$ – Nikos M. Sep 18 '14 at 22:55
13
$\begingroup$

It's an old (1965) and well-documented method called randomized response. It's used in various situations such as survey interviews or informal or ad hoc surveys conducted in classrooms or lecture halls. It's useful to think about the sample size that would be required to yield a given level of precision for the estimates one would obtain.

$\endgroup$
  • $\begingroup$ plus one should asume that the participants will actually give a truthful answer if given the random event that represents it. In the final analysis does not say much, one can still lie if one wants to. It is supposed to provide the capability to answer "almost anonymously" in a survey about issues that may be "criminalised", involve polarised or under-represented social issues, etc.. $\endgroup$ – Nikos M. Sep 18 '14 at 22:50

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.