New analysis from me & @mandl suggests that preprints might have driven global discourse about #nCoV2019 (#COVID19) transmissibility prior to the publication of relevant peer-reviewed studies. Find our preprint here [ssrn.com/abstract=35366…] as well as an explainer thread below! 
@mandl This alone was a really interesting finding, but we took it a step further by comparing the results of the relevant preprint studies against the relevant peer-reviewed studies. To do this, we first collected the basic reproduction number (R_0) estimates presented by each study.
As y'all may recall, R_0 is a measure of transmissibility *potential*. It can be defined as the *average* number of individuals a new case may infect in a fully susceptible population. I posted a few explainers about this measure earlier, including this more technical one below.
We also collected methods & data specs for each relevant study, which can be found in our preprint. Notably, the presentation of R_0 ranges differed across studies (e.g. 95%CI, 95%CrI, etc.); moreover, different modeling approaches & data sources were used across studies too.
Nevertheless, we found that the range of R_0 estimates presented by preprints overlapped with those presented by peer-reviewed studies later down the line. Mean preprint estimates skewed higher on average than peer-reviewed estimates (R_0 = 3.61 & 2.54, respectively). However...
These differences were driven primarily by two mean estimates that fell outside of the 95%CI we calculated for the preprint group's collective mean estimates [95%CI = 2.77, 4.45]. When these two estimates were discarded, average R_0 = 3.02 – similar to the peer-reviewed average.
I'm sure I'll have plenty more to say about this work later (all of which I'll add to this thread as appropriate), but I'm gonna leave it here for now... And as a reminder, our analysis has not yet been peer-reviewed & should thus be treated as *strictly provisional*. Thank you!
