* Initially it looked like maybe smoking was a notable risk factor.
* Then some data from a different population seemed to indicate maybe smoking was protective.
* Then further analysis indicated that all this data was likely skewed by upstream sampling bias (like pretty much all CV19 data so far).
* The latest speculation is that maybe smoking makes one less likely to catch CV19 but if you've are a heavy smoker currently (or recently), once you have CV19 your outlook is worse (but nowhere near as bad as being over 60 or immuno-compromised). This was based on a more recent study which seemed to have better data because it went narrow and deep on a few cases, however, it has the disadvantages of small sample size.
* My best guess is that the last bullet might be directionally correct but effect size is small enough that it could easily be swamped by other factors.
In general, us "armchair analysts" obsessively scraping the raw data sources and modeling on it quickly realize that all of the data is very noisy since different regions are using different criteria for what counts as a "case", "severe" and even "hospitalization". Plus no country is sampling the population in a consistent way because they have different criteria to get a test (which can also change daily and by area). For example, Italy's numbers look really troubling in terms of severity and progression until you look at the median ages of who they are testing, which skews 13 years older than their overall population.
I was curious if anything comes out of it, but I guess there's still nothing significant.
Source? I thought the contrary.