Right. I use the example that the number of circulating guns in the USA always increases.

So you pick a ten year period where per-capita homicide rate went up or down to 'prove' whatever correlation you want about guns and homicide.

Just reading the title reminded me of https://i.redd.it/4h8emlr5z3c41.jpg

Well, internet use -> computer use -> better economics + indoors more often.

Contrived but not ridiculous

There is no causation here: https://blogs.oregonstate.edu/econ439/2014/02/03/murder-rate...

More importantly I completely misread the trend of the graph i.e. for my hypothesis to hold the internet mark share should be increasing!

Must've wrote that half asleep or something

Read the paper maybe? There is a clear regime shift in the suicide data. Maybe there is some other trigger. But the argument of the paper is not without merit.

[1] https://www.ncbi.nlm.nih.gov/pmc/articles/PMC8159839/

I don't agree with this paper's statistical analysis, and I think there's probably a different trigger for the change BUT I came here to say that I strongly agree with you that the paper has to be taken on its specific merits. It's perfectly possible to have good epidemiology based on time series data.

(I feel like I should cite something. It's been a long time since I was a professional epidemiologist and I'm way too lazy to look up anything recent, but Oli Miettinen's book "Theoretical Epidemiology" is an example of how some epidemiologists are very clever and very careful.)

What? Are you telling me I can’t make a paper out of https://www.tylervigen.com/spurious-correlations ?