When people attribute your correlation to a selection effect you have to pull a switcheroo and claim if it wasn't for another selection effect the apparent effect would be even greater