which gives you whatever the necessary sample size is to make useful projections about the entire population. However, as soon as that sample is not randomized it doesn't tell you much about the total population (because of the selection bias you mentioned). As discussed last week, that's probably what these charts tell us more than anything, the selection parameters have changed over the past month as it takes less "evidence" to get a test.

I could also understand this being a testing capacity thing whether it's to deal with a backlog or to be prepared if things get worse.