Thread · 3 tweets · 22 Aug 2019

You are running an A/B test of tracking vs no tracking in a world in which 1) buyers can choose what to bid on and 2) the bidding architecture is centered on tracking. You're using that to draw inferences about a world in which tracking does not exist at all or barely.

I *really* would like to have a solid conversation on this, but if you don't see the problem with this approach you're missing the basic economics of the situation which we have to deal with.

You can make that finding as statistically meaningful as you want it will still be completely unrelated to the statement it is brought in to support. You're measuring unemployment in Lichtenstein with the LHC.