They don't say how they found these parents. Could be a biased sample. If it was an internet survey, where they posted the survey could determine what results they found - eg if they posted it on any GFCF forums, those forums would attract parents who think GFCF helped their child.
Plus, even if it wasn't biased, it's looking at parent impressions, and I doubt most of those parents have done any blinded analysis of this issue. If the parent knows they've cut out gluten and casein and they think it'll help, they might treat the kid differently, and this change in behaviour might be producing the positive results instead of the diet itself. To test that, you'd need to feed the kid food that may or may not have gluten and/or casein, and make sure neither the kid nor any of the adults interacting with them know what the food is. Then record the kid's behaviour, and see if the kid acted any different on gluten/casein versus GFCF.