2600 bro wrote:
Basically if the correlation between acetominophen and autism goes away when you compare siblings but exists when you compare all kids - you can conclude there is something inherent to families that both causes changes in acetominophen use and as well changes in chance of autism diagnosis. Sibling pairs are used a way to remove a large portion of the millions of effects that being raised in different families brings.
Basically its saying something(s) else is causing women to take more tylenol and causing an increased likelihood of their kid to be diagnosed with autism.
Consider a dumb, fake example:
Poor women take more tylenol, poor women also are more likely to have a kid diagnosed with autism. You can see how being poor probably causes millions of things that could relate to autism but it also makes tylenol consumption correlate with autism. However if you only compare siblings then you are only comparing poor kids to poor kids and rich kids to rich kids, so you've removed most of the effect of being poor from the analysis.
This is over-simplified but it starts to get at the purpose.
Reported for using too much common sense and logic on LRC