Read on Twitter
We've used our #Scotland #SIR model to interpret most recent sero tests published a few hours ago by @EEID_oxford. We propose a workflow to assess "rho" one of the biggest unknown of the #Covid19 story. We find a stunning low number rho=3e-3. Super good news ? [1/...]
2/ Data are here, published 4 hours ago https://twitter.com/EEID_oxford/status/1248662224010391553?s=20
3/ Data compiled from the original infographics, presented in a table. From blood tests, 0% sero+ on 17/3, and 1.2% of population at mid date 22/3. We have no time for more detailed stats... 
4/ From there, we estimate around 65,000 people in #Scotland were seroconverted #SarsCov2 on 22 March (maybe me?), a v. large number ! (On 17 March the number was NOT DETECTED from 500 blood tests, that's important).
5/ From https://doi.org/10.1101/2020.03.23.20041707, Lou et al, a #SARSCoV2 infected person is seroconverted in about 12 days (quicklook). From our #Scotland #SIR model we calculated the number of seroconverted persons by N*z(t-12). We ran simulation to match observed number 65,000 by varying rho
6/ Simulation results: curve N * z(t-12 days) with N #Scotland population, z = simulated infected + recovered + dead proportion.
7 / Simulation results: table for simulated seroconverted, showing match of 65,000 on 22 March. Interestingly, simulated proportion of seroconverted people on 17 March is only 0.46% explaining that the 17 March tests DID NOT RETURN A POSITIVE. Well timed samples ! 
8/ Full simulation results, with death rate grossly matched. We estimate 40% of population immunised at end of outbreak, 60% being the required threshold for full group immunity.
9/ General conclusion: rho=0.003 or 0.3% is the value obtained from this study. On the lowest side, as per @UniofOxford and @LourencoJML intuitions ? Lots work from academia to confirm, but we believe workflow presented here has some future.
10 /10. Uncertainties a) need new serotests a week after - high sensitivity to the "1.2%". b) We'd like our "t-12" peer reviewed c) all other model uncertainties previously published d) @CovModel to do same studies on other parts of the world and confirm if possible
