pgmj.bsky.social
PhD & lic. psychologist. Research and applied work at RISE Research Institutes of Sweden & Karolinska Institutet. R package for Rasch psychometrics: pgmj.github.io/easyRasch/
#openscience, #prevention, #psychometrics, #rstats, #photo
210 posts
2,222 followers
1,123 following
Regular Contributor
Active Commenter
comment in response to
post
Folks, always share your code. It doesn’t have to be perfect to be helpful. And if you feel that it’s still too messy or not sufficiently clean to be shared, you shouldn’t submit yet. After all, there could be mistakes in your mess.
comment in response to
post
Thanks for alerting me, this looks like a nice initiative! I'll look closer at the links later on. Would be great to see DFI available in JASP.
comment in response to
post
If you were asking specifically about the dynamic fit index issue, here are 3 earlier papers:
McNeish, D. (2023). doi.org/10.1037/amp0...
McNeish, D. (2023). doi.org/10.1080/0027...
McNeish, D., & Wolf, M. G. (2021). doi.org/10.1037/met0...
comment in response to
post
Indeed. I have the Fuji GS645S for shooting "half frame" 6x9 vertically without turning the camera on the side ;)
comment in response to
post
I have made materials introducing Rasch/IRT as well as a fairly extensive vignette to Rasch analysis using the R package I have developed (easyRasch): pgmj.github.io/raschrvignet... and github.com/pgmj/RaschIR... but I think neither works quite well without prior knowledge or handholding by someone
comment in response to
post
...since it needs revising (ongoing) to incorporate correct information about critical values for GOF metrics: osf.io/preprints/os... We basically use the 4 criteria as presented by Kreiner (2007), who refers to Rosenbaum (1989).
comment in response to
post
I think there is a lack of accessible resources. I like the book "Rasch models in health" (2013) and openpublishing.library.umass.edu/pare/article... I have an old preprint that tries to simplify and pinpoint key issues to look for in a paper but cannot recommend it in its current state...
comment in response to
post
Ooops, forgot to paste the final point... If you have any of the issues listed (except local dependency issues), you will not be able to make reasonable comparisons to other research results at all, even if everyone uses the same questionnaire.
6/6
comment in response to
post
If you read this far, please help spread the knowledge that we need to stop using rule-of-thumb critical values for model fit metrics (Hu & Bentler, Kline, Brown, Bond & Fox, etc). There are methods available for adaptive thresholds, see doi.org/10.1080/1070... and pgmj.github.io/rasch_itemfit/
comment in response to
post
If response category thresholds are disordered (almost always seen when using too many response categories in combination with no labels or only labels on endpoints), you will have a lot of noise in your data, making comparisons even harder. Also, you are likely to have issues with invariance. 5/6
comment in response to
post
If there is local dependency between items (items should only be related through the latent variable), you will see inflated change scores (if there is change). 4/6
comment in response to
post
If invariance has not been adequately tested/proven (across gender/age/time/etc) the measure might not be suitable for making comparisons at all, except perhaps for within certain groups. 3/6
comment in response to
post
If dimensionality has not been adequately tested/proven, you are not measuring a coherent latent construct, maybe not you are not even measuring what you are intending to measure. (disclaimer: well-functioning dimensionality does not guarantee content validity either) 2/6
comment in response to
post
I expect that the target group for this is those who got into X100 cameras based on influencers recommending it.
comment in response to
post
Take one look at how climate scientists have tried to ensure their measurements taken over time are comparable, to refine measurement of temperature, to cross-verify using related records (tree rings, ice cores etc). What does psychology/psychiatry have?
comment in response to
post
Dags för lite semester?
comment in response to
post
It's bonkers that in academic science we spend hours obsessing over the wording of a paper, but often only one person has seen the code that produced the results! 2/N
comment in response to
post
Exakt detta. Det verkar riskfritt som företag att testa olika strategier för att maximera intäkter. Värsta konsekvensen är återbetalning.
comment in response to
post
I wrote a bit about this 5 years ago: medium.com/@pgmjoh/abou... and have since then co-authored some related papers: www.sciencedirect.com/science/arti... and link.springer.com/article/10.1...
comment in response to
post
@andrew.heiss.phd has a great overview of the pros and cons of #rstats table packages 👇
bsky.app/profile/andr...
comment in response to
post
This follow up paper on heterogeneity is also great: vuorre.com/heterogeneit...
comment in response to
post
Snyggt med rektangulärt, men det där ser ut att vara för betydligt mera vidvinkligt objektiv? Jämför med squarehood.se/products/squ...
comment in response to
post
Väldigt trevlig kamera, har också haft en. Finns en del fantastiska objektiv till den. Rekommenderar varmt 35/1.4 och 56/1.2 som båda är exceptionella. 14/2.8 är också väldigt bra om man gillar vidvinkel.
comment in response to
post
Maybe even shorter using ifelse()? :)