+Gert Sønderby How good is your wife with statistics? I realized after , If I was right ( One-size-does-not-fit-all ) traditional experiments might not be good enough, to see how and who is benefiting from interventions.
A lack of control against background results is a joke. But if it helped and harmed different people, it might look averaged out.
You would need to test sstatisticalcorrelations for change-itself.
This is a similar problem for antidepressants
( may help a subset population but doesnt beat placebos long-term on a broad population )
One way to establish this is, with re-challenge studies and re-iterative testing on subset populations that show statistical significant correlations. On a drug you want to show you can turn on and off an effect.
If done correctly this is a valid way to show a subset can get results that beat placebos.
I dont know how to apply this idea to Autism-interventions. This seems to be a form of cherry-picking, but seems valid if you can result in a full group that gets results. They you can test for why.
Its a Complexity-problem.
I am also wondering what sort of sham treatments it would take to test the Acting classes.