Here is the paper discussed (Buss et al. 2021) -- https://science.sciencemag.org/content/371/6526/288?ijkey=0164586a5cf47a56724f873706982763e64356b8&keytype2=tf

Here is the paper discussed (Buss et al. 2021) -- https://science.sciencemag.org/content/371/6526/288?ijkey=0164586a5cf47a56724f873706982763e64356b8&keytype2=tf_ipsecsha https://science.sciencemag.org/content/371/6526/288?ijkey=01... The October 76% attack rate figure for Manaus seems to rest on the assumptions that (1) blood donors are more likely to be asymptomatic and (2) asymptomatic infections are more likely to be "serosilent" (ie not visible via IgG antibodies). (1) seems likely a straightforward assumption, but (2) doesn't seem obvious to me or to be backed up by the reference (https://pubmed.ncbi.nlm.nih.gov/32793928/?dopt=Abstract https://pubmed.ncbi.nlm.nih.gov/32793928/?dopt=Abstract). Since lot of the analysis of the viability of herd immunity as a strategy in the Science article rests on Buss et al's estimation of the attack rate (and by extension the estimation of the seroconversion rate in asymptomatic cases), it seems important to get right. Which is to say, if the estimation of attack rate is significantly too high, then Manaus might not have yet approached the HIT, and the whole analysis is wrong. Can someone with more background here explain (1) the methodology of estimating the seroconversion/seroreversion rates in asymptomatic cases, (2) whether asymptomatic cases are more likely to be "serosilent", and (3) anything else I might be missing/misunderstanding? Regarding the seroreversion estimation methodology: it seems like they simultaneously reduced the threshold for what counts as a positive and modelled seroreversion via exponential decay: "[W]e assumed that the probability of an individual seroreverting exactly m months after recovery decays exponentially with m. We estimated the decay rate and the proportion of patients who seroreverted using the seroprevalence data from Manaus to find the decay rate that minimized the number of new cases in July and August while avoiding decreases in prevalence". Is that a standard way to model seroreversion? This seems to model it quite differently: https://immunology.sciencemag.org/content/5/52/eabe0367 https://immunology.sciencemag.org/content/5/52/eabe0367