writersoul:
I think ben is trying to say that sometimes a statistical study will be way wrong, and if it is obviously wrong, the intelligent thing to do is to ignore it.
We generally consider something statistically significant if the results are 2 standard deviations off the mean assuming the null hypothesis. Yet, 5% of surveys will yield a result that is 2 standard deviations off the mean! So 5% of them are wrong.
Usually, we assume we are in the 95% which are correct. But if I see with my own eyes that the results are wrong, I won’t.
Here is my example:
I have 100 children, and I think that there are 50 boys and 50 girls (big family, I’m mormon). But just to make sure, I do a survey, and randomly ask 20 children if they are a boy or a girl. All 20 are girls. Do I conclude that all my children are girls?