In a way, it’s sad this amazing read by @adambecker.bsky.social had to be written. But it’s an incredible debunking of the alleged science behind what people like Bezos and Musk tell us the future will be like. But it’s also more than that…

Is it still a descriptive study if you are comparing a measure across groups (e.g. differences or ratios)? Or are descriptive studies only univariate measures like prevalence and incidence?

Here, the latter is claimed:

Can't stop thinking about these lines from @economeager.bsky.social in rottenandgood.substack.com/p/taking-our...

A perfect description of, let's say, sub-optimal behaviour I sometimes observe or hear about. They "know" the right answer, but sometimes the data need help to "know" it too.

Oxford Dictionary of Statistics

Association: Two variables are associated if they are not independent, i.e. if the value of one variable AFFECTS the value, or the distribution of the values, of the other.

AFFECTS?! (said in an outraged Basil Fawlty voice)

My middle name is Alexander. My two given names are two of the names to lose the most popularity since 2000 in Quebec.

I blame Pearl Jam.

For causal inference people who have never encountered this, it's a nice little thing to think through. How would you answer this? Does Suzy's rock cause the bottle to break? It would have broken anyway...

From: plato.stanford.edu/entries/caus...

Does it though?

Results: physical activity was associated with

Conclusions: physical activity may be associated with

The word associated cannot mean the same thing in both sentences.

This is the appeal of the word associated, it can mean anything you want and no one is going to ask you what you actually mean.

Dubrovnik solved the trolley problem actually

Unbelievable

JAMA's policy used to be that observational research couldn't use causal language but you could with mediation. They gave an example where you need at least (maybe many more) nine times as many exchangeability assumptions as a regular exposure-outcome analysis. med.stanford.edu/content/dam/...

This type of error is everywhere and it’s like the third day of stats 101. The penalty for making this error should be 1 year no publishing and you have to retake two semesters of stats.

For those who still remember Wayne's World

This is a thing I read. And the "more on this below" does not help.