Option 1 and label the old ones as deprecated 🤔 ? FAUX just looks nice and snazzy on the hex sticker.

Some already do, and it's quite a trip trying to write a plain-language summary about heteroscedasticity.

This post went from "Oh fun!" to heart palpitations really fast.

I often end up using super long object names because a_meaningful_name_is_always_better, we have the tab-autocomplete, it's *fine*.

The idea that we have to "encourage responsible AI usage" is a core part of the inevitability narrative driving the institutional adoption that makes people think it's worth trying out.

it contributes to the very problem it supposedly tries to solve. we can't let it go unchallenged.

I did hear a bit doing the traditional "Open access publishing is now so expensive no-one can afford to do it". And I mostly want to stand up and scream Green.

I seem to spend my life wanting to stand up and scream green.

Glad someone found a use for it(: the accepted version has the same key messages but includes a more rigorous literature search strategy and clearer technical definitions.

In psychology, CIs are pretty common so I guess it depends on the discipline.

I love raincloud plots - it's a nice way to combine raw data distributions + summary stats in one plot. Here's an R package with some examples: github.com/njudd/ggrain

I'd use confidence intervals because they're more useful for inference 🤷. But if both SD and SEM are relevant and needed, why not plot both? You could try something like a raincloud plot with the SD displayed on the "cloud" and SEM around the mean.

I do descriptive stats first with psych students - show what kind of measures can be used to summarise data (central tendency and dispersion), then show how to plot them and how _not to_ plot them.

Just above The Long Man of Wilmington (walked from Eastbourne). Devil's Dyke is on the list next time we get weather like this.

As someone who barely posts, I've narrowed it down to this: bsky.app/profile/prof...