I had been struggling for a while to get my head around the argument, made by @BillHanage, @NateSilver538 and others on this website today, that antibody test results from places with low infection rates will be much less reliable than those from places with high rates 1/5

So yesterday I tried the Gerd Gigerenzer et. al. approach of converting percentages to numbers, which is described in the article below. It worked! So here goes ... 2/5 https://www.psychologicalscience.org/journals/pspi/pspi_8_2_article.pdf">https://www.psychologicalscience.org/journals/...

There& #39;s a population of 100 people and you give them a test with a (very high) specificity of 99%, meaning that it gives a false positive result only 1% of the time. So in your population of 100, you can expect to get 1 false positive. 3/5

If 20 of the 100 people are in fact infected, you& #39;ll get 21 positives. Not a big deal! If only 1 in 100 are infected, though, you& #39;ll get 2 positives, making you think the infection rate is twice what it really is. 4/5

So whether the true infection rate in Santa Clara County is 1% or 2% or 3% or 4% of the population (I& #39;m guessing 1%, but that& #39;s just me), any study aimed at finding out what it is will be less reliable than one conducted where the infection rate is much higher. Like, say, NYC 5/5