[darawk]

It is extremely far from as certain as science ever is. Randomized, controlled experiments are as certain as science ever is. This is pretty far from that.

The evidence is as good as any non-controlled experiment is going to be, perhaps. But it's pretty important that we not confuse these kinds of descriptive results with the actual certainty that comes from high powered controlled trials. It's extremely easy to make reasoning errors in descriptive data analysis, that look very convincing.