Forums:
Well, the people were bad too... but that's beside the point.
Hey, let's look at data from a single website... and draw nationwide conclusions from that!
https://www.theguardian.com/science/2018/jul/19/ashley-madison-democrats...
Where do we even start?
One website. Maybe the others prefer a different site?
One mechanism. Maybe the others don't use the internet for that?
One vector. Maybe the others don't need to hide it... everyone knows they're doing it anyway?
But let's just assume everything. No need to hold other variables constant...
Bad science.