it's a worthwhile point. i considered that issue a ways back and looked for ways to tease out the selection issue. what seemed to emerge was that above a certain sample rate (that's very low vs current levels, maybe 250k vs 1-1.5 million now) you can find a bit of evidence, but not at higher rates. it seems to saturate and vanish as an issue. so we saw it in the positivity figures early on (likely because most testing was among symptomatic) but it more or less vanished by april 2020. the testing numbers here have been so high for so long that selection bias looks to have washed out pretty quickly.