Meno,
Ok keej, here�s a hypothetical study. Let�s see what you think:
The purpose of the study is to determine the risk factor in the general population for autism.
I take a random sample of 100,000 people.
I remove 10,000 people who are wearing red socks, because I hate red socks. (I�m justified in doing this, because we can all agree that the color of one�s socks has no effect on his risk for autism, so the autism ratio in the remaining group should be the same as in the original group.)
I�m now left with 90,000 people. I find that 900 have autism. I conclude that the risk factor for autism among the general population is 1:100.
Thoughts?
You said you’re justified in removing the 10,000 with red socks. I’m not so sure. What if coincidentally, there was a concentration of either Neuro-typical or Autistic children in that group. That would make a very big difference on the resulting ratio.
In my view at this point in time, the only justification of removing anyone is if they weren’t subject to the initial risk.