Thread by @Ayjchan

The coverage of the newest preprints by the Proximal Origin authors and friends must be breaking some kind of record.

Has anyone else heard of preprints being featured on @nytimes front-page news? Not to mention at the same time as a war is starting.
nytimes.com/interactive/20…

Glad to see that many of the @nytimes subscribers who commented on the article have solid critical thinking abilities.

To quote one of them:
"I wish @carlzimmer had taken the time to do more research before writing this article."

@nytimes @carlzimmer It's great that this @nytimes front-page news article is being proactively archived at different times to show its evolution as @carlzimmer appears to be reading the work and updating his article with figures and pictures from the preprints.
archive.fo/https://www.ny…

First version of "New Research Points to Wuhan Market as Pandemic Origin" in the @nytimes has barely any info except for the length of the preprints, their final conclusions, and one quote from an author.
archive.fo/SmvK5

@nytimes I'm not sure how much time @carlzimmer had to review these preprints before they went public. One question I think should be put to the authors is if they can represent the sampling data more accurately by showing the number of samples taken per stall...

https://twitter.com/Drinkwater5Reed/status/1497822789704634371

The figure currently being featured in the @nytimes only shows the locational distribution of the number of positive samples. However, even elementary school children should know to ask for the denominator.

We know that sampling was not even across the market.

@nytimes I hope it makes sense that if you sample one stall 100 times, you shouldn't be shocked to find more positives there compared to another stall that you only sampled 10 times.

@nytimes I'm a bit surprised that this basic principle failed to be grasped by the eminent scientists on the authorship of the preprint. One can only hope that peer reviewers at top journals will catch this.
zenodo.org/record/6299116…

@nytimes If a high school or undergraduate researcher had produced this figure for me, I would've asked them immediately to go back and remake it after normalizing for the number of samples taken per stall.

@nytimes And if the researcher told me that they don't know how many samples were taken per stall, then I would tell them that they cannot make a figure like this.

It would be ultimately misleading and unscientific to present incomplete data.

To the @nytimes journalists covering this story @carlzimmer @benjmueller you don't need a PhD in anything to understand that the way the positive samples are depicted without information about sampling effort across the market can be deeply misleading.

@nytimes @carlzimmer @benjmueller I'm still working through these preprints, but another problem is that their maps don't point out the location of the Wuhan CDC and the main Wuhan hospitals. One could easily argue that these were the epicenter of the outbreak as opposed to the Huanan market.

Map from my book:

@nytimes @carlzimmer @benjmueller The December 2019 epicenter of the Covid-19 outbreak was the center of Wuhan city where all the metro lines intersect, the main hospitals are located, and there is the highest population density and concentration of elderly residents.

I hope this is not surprising to anyone.

@nytimes @carlzimmer @benjmueller In case there is confusion, Fig 4C shows distribution of (+) env samples, but 4D shows distribution of stalls sampled, *not* the distribution of samples taken. I went through their supplementary tables but there is no data on total samples taken per stall.

@nytimes @carlzimmer @benjmueller I also think @nytimes @carlzimmer @benjmueller that your reporting should raise the possible conflicts of interest among the authorship.

The study was conceived by Kristian Andersen of Proximal Origin/Feb 1 call fame, and the project was funded in whole or in part by NIAID.

@nytimes @carlzimmer @benjmueller The authors even write that they cannot authenticate the CCDC report which their analysis is based on. They don't know the sampling scheme across the market. But they somehow note that their results are "robust to significant sampling biases".

How do they know this?

@nytimes @carlzimmer @benjmueller Is this a thing in science?

We don't know if the data is correct. We don't actually know the data. But we're sure that our analysis is robust to biases that we don't know about.

I guess these preprints will be published at @ScienceMagazine @CellCellPress or @Nature in a few weeks.

@nytimes @carlzimmer @benjmueller People are tweeting at me to remind that these preprints ultimately fail to identify an intermediate host for SARS-CoV-2 or any direct evidence of a natural spillover at the market.

In other words, the evidence for their hypothesis is as strong as that for a lab #OriginOfCovid

@nytimes @carlzimmer @benjmueller I hope these scientists will use the term "dispositive evidence" more judiciously. Public trust in science is at an all time low when we need it to be at an all time high, and it's not helpful to say that evidence is dispositive when it is only circumstantial.

@nytimes @carlzimmer @benjmueller The same applies to science reporters. If the public sees you promoting incomplete research, for which data is missing or assumed, as "a significant salvo" in a controversial issue, it will shake confidence in our science communication and journalism system.