1/
Interesting study coming out suggesting full lockdowns offer no benefit over softer measures.
Great they released the code! It is worth a deeper dig as the journal is fairly vague about the data used. I've stuck the source code on my GitHub here: github.com/dr-harper/NPI-…
Interesting study coming out suggesting full lockdowns offer no benefit over softer measures.
Great they released the code! It is worth a deeper dig as the journal is fairly vague about the data used. I've stuck the source code on my GitHub here: github.com/dr-harper/NPI-…
https://twitter.com/RealJoelSmalley/status/1348226804851613696
2/
It might sound fair to accuse me of bias, with the only reason that I question this study because it challenges my view of lockdowns.
I have seen papers suggest lockdowns don't have a big impact, but this seems to suggest they have a negative impact? What is going on
It might sound fair to accuse me of bias, with the only reason that I question this study because it challenges my view of lockdowns.
I have seen papers suggest lockdowns don't have a big impact, but this seems to suggest they have a negative impact? What is going on
3/
Now, one threat not considered by the paper is potential confounding variables. See statisticshowto.com/experimental-d…
Is it fair to compare international countries purely on case rates and lockdown measures? Are there broader differences? Highly possible
Now, one threat not considered by the paper is potential confounding variables. See statisticshowto.com/experimental-d…
Is it fair to compare international countries purely on case rates and lockdown measures? Are there broader differences? Highly possible
4/
So lets look at the code!
It is worth noting that for peer-reviewed journals, reviewing the code is normally optional!
This job is made harder when the code is confusing. So instantly my alarm bells are off when I am seeing a mix of R and Stata.
So lets look at the code!
It is worth noting that for peer-reviewed journals, reviewing the code is normally optional!
This job is made harder when the code is confusing. So instantly my alarm bells are off when I am seeing a mix of R and Stata.
5/
The journal provides fairly minimal information about the data. Specifically, no timespan is provided. Looking at this we have fairly different temporal temporal coverage between February - September. Unclear how much was used for the modelling.
The journal provides fairly minimal information about the data. Specifically, no timespan is provided. Looking at this we have fairly different temporal temporal coverage between February - September. Unclear how much was used for the modelling.
6/
I'd love to see if others have any opinions on this! Unfortunately, my knowledge of #STATA is non-existant, so I can't be too detailed on the modelling side of things
🙁 feel free to fork the code and play around! @VictimOfMaths
I'd love to see if others have any opinions on this! Unfortunately, my knowledge of #STATA is non-existant, so I can't be too detailed on the modelling side of things
🙁 feel free to fork the code and play around! @VictimOfMaths
7/
Don't get me wrong. Peer-reviewed journals are definitely a gold standard when it comes to evidence, so I wouldn't want suggest otherwise. But given the potential for interpretation in this case, it is crucial to be critical.
Don't get me wrong. Peer-reviewed journals are definitely a gold standard when it comes to evidence, so I wouldn't want suggest otherwise. But given the potential for interpretation in this case, it is crucial to be critical.
8/
Just adding some other analysis which is 100 times better than mine!
Just adding some other analysis which is 100 times better than mine!
https://twitter.com/GidMK/status/1349164532627693570
• • •
Missing some Tweet in this thread? You can try to
force a refresh
