New blog post in which I discuss measurement invariance from a causal perspective. Come for the amateurish featured image (who needs photoshop anyway), stay for fancy Pooh as figure label. Bonus: You'll finally understand the different levels of invariance (I hope). www.the100.ci/2024/01/10/a... - ThreadSky

dingdingpeng.the100.ci • 438 days ago

New blog post in which I discuss measurement invariance from a causal perspective. Come for the amateurish featured image (who needs photoshop anyway), stay for fancy Pooh as figure label. Bonus: You'll finally understand the different levels of invariance (I hope).
https://www.the100.ci/2024/01/10/a-casual-but-causal-take-on-measurement-invariance/

Comments

bwiernik.bsky.social•438 days ago

Very nice post!

My favorite example of residual non-invariance is something like gender or country differences in use of extreme vs middle response scale options

michaelmittag.bsky.social•438 days ago

Just realized my post below is pretty much another example of your point here. Absolutely!

dingdingpeng.the100.ci•438 days ago

Oh that’s an interesting one that I hadn’t considered before!

michaelmittag.bsky.social•438 days ago

Learnings: 1) You can compare factor solutions using Tucker's congruence coefficient after performing procrustian rotation (I didn't know that even existed). 2) Always check the source items for easy explanations. Eg Beck Depression Scale is also affected just if sleep improves.

michaelmittag.bsky.social•438 days ago

Hey cool! I just realized I co-wrote a paper on that (comparing creativity in eastern and western societies). In our case, the issue was supremely simple: The scale gave higher scores the more items you checked, so it may just measure how much different cultures like checking survey items. 🤷‍♂️

michelnivard.bsky.social•438 days ago

Very cool! (Also, not entirely unrelated to the Eronen paper that just came out, I guess it’s measurement week…)

dingdingpeng.the100.ci•438 days ago

I had finished this before Christmas and seeing that people are discussing measurement again gave me the necessary momentum to coerce my cobloggers into providing feedback

jamiereillycog.bsky.social•438 days ago

LOVE!

drewhalbailey.bsky.social•437 days ago

This is great, and great to see @protzko.bsky.social getting a shout-out; I think he was ahead of the curve on this!

tontrug.bsky.social•438 days ago

Nice and helpful, thank you. Would be interesting to know how close this is to the epi term "undifferential measurement error" and the implication this property has on the direction of bias. https://link.springer.com/book/10.1007/978-3-030-82673-4

tontrug.bsky.social•438 days ago

Undifferential misclassification to be precise (epidemiology is mostly concerned with categorical variables)

dingdingpeng.the100.ci•438 days ago

That would indeed be interesting, I haven’t read up on causal models of error in epi yet (but I probably should)

dingdingpeng.the100.ci•438 days ago

(Very disappointed to see that ProfMattFox is not yet on bsky 🥲)

tontrug.bsky.social•438 days ago

He used to be very active on Twitter, at least when I used to be there.

tedmond.bsky.social•436 days ago

Dang Julia I'm always impressed by your ability to make scary concepts feel approachable. Especially because this post tacitly supports my opinion that Metric/Scalar are The Fun Ones For Theory and that config/residual are booooooring hangers-on

dingdingpeng.the100.ci•436 days ago

Thank you Ted 😊 glad I could support your prejudices 😂

lornejcampbell.bsky.social•437 days ago

Timely and helpful for us. Grad student Omar Camanto has been teaching me all about measurement invariance and we have a paper using it almost ready to send out. Always appreciate your papers and posts—thank you.

drrodebaugh.bsky.social•438 days ago

A better introduction to measurement invariance than any of the others I have read! Make this paper now? Bitte? (you don't have better things to do do you? 🙃

Comments

Posting Rules

Reply