> Disease spread tends to always show a pretty standard shape. It was that observation (Farr's law) that led to epidemiology being born as a field. It grows, it

> Disease spread tends to always show a pretty standard shape. It was that observation (Farr's law) that led to epidemiology being born as a field. It grows, it peaks, it enters immediate decline. Only without intervention. With social change, the disease's R0 (with respect to a given society) will differ, altering the shape of the epidemic curve. We can have great confidence that COVID has not gone through a full, "status quo" epidemic curve anywhere. Take Italy as an example. That nation has conducted (as of May 2, Wikipedia data) 2.11 million tests to find 210k positive cases. That 10% test positivity rate forms a loose upper bound on the prevalence of nCoV in the entire population -- even if we assume that policymakers erred and many infections are asymptomatic, test-selection criteria should not have caused a worse than average chance of detecting a positive case. In the meantime, Italy documented 28,710 COVID-related deaths as of that date. If we again make the generous assumption that all COVID-related deaths were detected but the true population prevalence was about 10%, that would give the disease an 0.48% IFR. That's far too high for a rapidly-spreading disease to remain hidden for long. Simultaneously, we can't say that the supposed 10% infection rate is sufficient for herd immunity. If 10% of the population is infected now when the virus was introduced at the end of December, the R0 of the disease must be well above 2 (with a generously short two-week period between infection and recovery -- shorter than many observations -- we've only had 8 generations.) Herd immunity would then require > 50% immunity in the population. Instead, the much simpler conclusion is that policy responses have worked, with lockdowns and distancing reducing Italy's R number to somewhere around 0.75 (based on a rough look at the number of new cases per day, divided by the number from two weeks ago). > All we can do is work to piece together a narrative of what really happened by examining all the evidence we can get. Yes, but we must examine all that evidence in light of what we know of epidemiology. It's far too easy to cherry-pick data that is comforting or aligns with our political predispositions. > Repeated anecdotes from many different people about having had a COVID-like illness before it was being discussed much are interesting for that reason. ... but the plural of anecdote is not data. Especially for a disease such as COVID, where the range of attributed symptoms is so wide that just about any commonly-circulating cold or flu could -- by symptoms alone -- be attributed to nCoV.