|
William McAuliffe
@
w_mcauliffe
Boston, MA
|
|
3x alum @univmiami, now a postdoc @harvardmed. A psychologist passionate about arepas, pad kee mao, and coffee.
|
|
|
512
Tweetovi
|
137
Pratim
|
186
Osobe koje vas prate
|
| Tweetovi |
|
William McAuliffe
@w_mcauliffe
|
8 h |
|
😥
|
||
|
|
||
|
William McAuliffe
@w_mcauliffe
|
19 h |
|
|
||
|
William McAuliffe
@w_mcauliffe
|
2. velj |
|
Invariably the longer I have sat on a paper the better it has become due to better understanding of methods/stats considerations that are in my control. It can be frustrating to publish something and right away able to make a long mental list of how I would do it differently now.
|
||
|
|
||
|
William McAuliffe
@w_mcauliffe
|
2. velj |
|
finally, also agree funder's point that much statistical advice is given with a heavy dose of moralistic contempt that is knee-jerk and serves nobody well.
from following methods discussions though i have had revelations that have constructively changed the way i do things.
|
||
|
|
||
|
William McAuliffe
@w_mcauliffe
|
2. velj |
|
Could also be that some assumptions are just less relevant to your work. If studying evoked culture the non-invariance is kind of the point. If you're using a measure for hiring/admission, easy to understand why ETS invests in findings invariant items.
|
||
|
|
||
|
William McAuliffe
@w_mcauliffe
|
2. velj |
|
If you mean that measurement violations often don't affect qualitative conclusions, I agree and you can demonstrate using sensitivity analysis, DMACs, etc.
If skeptical about validity of measures, then makes sense to be especially reluctant about latent comparisons.
|
||
|
|
||
|
William McAuliffe
@w_mcauliffe
|
2. velj |
|
Agreed. I used it b/c I think it is a very intuitive, if extreme, example of why invariance is important.
I doubt they would have IRT invariance, but I agree that you can contest comparison on non-stats grounds. E.g., we would both resist B5 compar. b/c we think model is wrong.
|
||
|
|
||
|
William McAuliffe
@w_mcauliffe
|
2. velj |
|
Once you have invariant indicator to set scale and location, you can include other non-invariant indicators.
Econ game was meant to be an ex. of non-invariant indicator that is useful for studying contingency. Just don't infer that UCLA student is more fair than machiguenga.
|
||
|
|
||
|
William McAuliffe
@w_mcauliffe
|
2. velj |
|
direct comparisons = group difference in latent means or magnitude of correlation of latent variable with other variables.
|
||
|
|
||
|
William McAuliffe
@w_mcauliffe
|
2. velj |
|
diagnosis indeed requires at least two invariant indicators. point was that you *don't* expect it of many items. but there are indefinitely many indicators, so finding two is possible and many design measures with an eye towards invariance for certain group comparison.
|
||
|
|
||
|
William McAuliffe
@w_mcauliffe
|
2. velj |
|
Many comparative studies aren't making latent claims. Econ game studies across cultures almost never do. But some data are made with the explicit intention of making a direct latent comparison. Those projects would benefit from thinking ahead of time about which measures to use
|
||
|
|
||
|
William McAuliffe
@w_mcauliffe
|
2. velj |
|
It's not an expectation that invariance var/covar will hold, just that some indicators are better than others for making direct comparisons. If yr making direct comparisons at latent level, at least quantifying the bias introduced by nonvar. is an inescapable requirement.
|
||
|
|
||
|
William McAuliffe
@w_mcauliffe
|
31. sij |
|
I actually quite liked that schedule.
|
||
|
|
||
| William McAuliffe proslijedio/la je tweet | ||
|
The Virtue Project
@bcvirtueproject
|
29. sij |
|
Check out this new feature piece in @bcheights highlighting @LianeLeeYoung & @kmcaulif1's work on "The Big Questions". bcheights.com/magazine/index…
|
||
|
|
||
| William McAuliffe proslijedio/la je tweet | ||
|
Hugo Mercier
@hugoreasoning
|
28. sij |
|
This time it IS out! Not Born Yesterday, arguing that people aren’t gullible. Psychology! Political science! History! Anthropology! Media studies! ToC below, and threads to follow with stuff I wish I’d put in the book
sites.google.com/site/hugomerci… twitter.com/PascalBoyerUSA…
|
||
|
|
||
| William McAuliffe proslijedio/la je tweet | ||
|
EJP
@EJPBlog
|
27. sij |
|
How does children's sociometric status relate to their personality? In a recent interview @vilmarine told us about his recent multi-method study on their cross-lagged associations. Read all about it on the #ejp blog at ejp-blog.com/blog/2020/1/27…
|
||
|
|
||
| William McAuliffe proslijedio/la je tweet | ||
|
Johannes Mahr
@JoBMahr
|
22. sij |
|
New article by @GCsibra and myself!
If you've ever wondered why the past plays such an important role in human (social) life, find out about our attempt at an answer here: doi.org/10.1177/174569… pic.twitter.com/QFBJnC1wzA
|
||
|
|
||
| William McAuliffe proslijedio/la je tweet | ||
|
Steve Stewart-Williams
@SteveStuWill
|
21. sij |
|
10 Things About Gullibility You Probably Don't Know. Great piece by Hugo Mercier, author of the forthcoming book Not Born Yesterday. press.princeton.edu/ideas/what-do-… HT @hugoreasoning @PrincetonUPress pic.twitter.com/2s4B9USXSS
|
||
|
|
||
|
William McAuliffe
@w_mcauliffe
|
23. sij |
|
We share your skepticism. Here is the main figure from the paper:
osf.io/2cn5d/
"RE" is the uncorrected estimate, whereas the other rows present estimates from techniques that each address publication bias in a slightly different way.
|
||
|
|
||
| William McAuliffe proslijedio/la je tweet | ||
|
Quillette
@Quillette
|
22. sij |
|
"The fact that we are better at suppressing our empathy than we are at boosting our empathy could suggest that we are walking around, out in the wild, with reasonably high amounts of empathy already."
quillette.com/2020/01/21/can…
|
||
|
|
||