1. An odd research study out of U. Manchester today uses an indirect and, frankly, bizarre method to estimate the incidence in the UK as being vastly higher than that inferred using more direct approaches. https://twitter.com/DrAdrianHeald/status/1260951024954638337 …
-
This Tweet is unavailable.Show this thread
-
2. The principal claim is that "unreported community infection may be >200 times higher than reported cases", meaning that "29% of the population may already have had the disease." (Most estimates from the US, EU, UK are closer to 10x than 200x) The tabloids are there:pic.twitter.com/hMtooxmxhk
5 replies 17 retweets 113 likesShow this thread -
3. Things get odd right from the very start. The first line of the paper's abstract is not your usual way of beginning a scientific report. https://onlinelibrary.wiley.com/doi/epdf/10.1111/ijcp.13528 …pic.twitter.com/RbQNVtE5P1
9 replies 29 retweets 282 likesShow this thread -
4. But let's get to the science. What did the authors do? They start by estimating a local R value that they call the Average Daily Infection Rate, and estimating its derivative. One could dig into this more deeply, but let's keep going instead.pic.twitter.com/dlGMOB9D43
3 replies 8 retweets 98 likesShow this thread -
5. What determines R? In an effort to estimate this, the authors use a regression approach across local regions. There's no underlying mechanistic model of how R changes with time, interventions, etc., nor any temporal analysis. Notice case density is from an April 8th snapshot.pic.twitter.com/YVjcp3Vsvh
2 replies 11 retweets 98 likesShow this thread -
6. Of the predictors in the regression, only cases/1000 people is predictive, albeit with an r^2 of 0.2. The authors then posit a linear relationship between R and case density: R_ADIR = 1.06 - 0.16 x Current Total Cases/1,000 population. Here's that data.pic.twitter.com/1XH9gz3XMQ
11 replies 7 retweets 90 likesShow this thread -
Replying to @CT_Bergstrom
Uh, surely this is exactly what you'd expect? They've calculated R from local increase in cases, which is also used to calculate cases/1000 so it's probably more surprising that they aren't more closely correlated
1 reply 0 retweets 2 likes -
Replying to @GidMK @CT_Bergstrom
Also, they've created a new "R" which seems to me to be very misleading because it's not at all the effective reproduction rate it's some odd unique calculation
1 reply 0 retweets 5 likes -
Replying to @GidMK
For most bad papers, I'd write the whole thread about that alone. Here it didn't even make the cut.
1 reply 0 retweets 6 likes -
Replying to @CT_Bergstrom
I get you, this is awful. But seriously, their R is quite clearly collinear with cases/1000 so it makes no sense to regress them at all. It's just bad maths if nothing else
2 replies 0 retweets 2 likes
Oh, also noticed - the extrapolation to "immunity" is clearly wrong. Even using their calculation and the rubbish in this paper, R(ADIR) = 0 has no bearing on immunity per se, simply on transmission. Could be due to social distancing!
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.