Right, and you have to rule out the "superior sponsorship" effect, above. I think you'd want to control for that by having a large sample size of unsponsored r

Right, and you have to rule out the "superior sponsorship" effect, above. I think you'd want to control for that by having a large sample size of unsponsored runner, who were accustomed to their chosen model of shoe, and then examine their race times vs shoe.