PBA (sorry, PBO): When I say a scientific survey, I assume that statistical analyses have been performed, such as standard deviation and chi square. (Trust me, I’m an ex-AP Bioer- I know of what I speak…)
A study such as one you describe would never pass these tests, as you point out.
benignuman: Statistical tests such as the ones describe above could be helpful in disregarding a statistic WITHOUT just doing so because it feels right.
Just as an illustration of what I mean-
Let’s say the American divorce rate is 50%. But wait- where I live, I know of very few people who are divorced or who have been divorced? That, however, is because I live in the Orthodox Jewish community, and there are fewer divorces.
That, as I mentioned, is what I don’t like about NASI- they lump all problems together, when specialization to uncover the roots of the problem may be a better idea.