search

philchalmers / mirt

Multidimensional item response theory
https://philchalmers.github.io/mirt/
201 stars 75 forks source link

Centering in estfun function #161

Closed RDebelak closed 5 years ago

RDebelak commented 5 years ago

The estfun function was updated to allow centering of the case-wise score contributions (i.e., setting their expected values over all persons to 0). This centering does not affect the results of this function in a maximum likelihood framework, but allows the application of score-based invariance tests in Bayesian framework.

philchalmers commented 5 years ago

Thanks. Can centering = FALSE be the default instead? It seems odd to me that centering the derivatives would be a default to this function.

RDebelak commented 5 years ago

Hi Phil,

that’s alright with me. My reasoning was that this function was initially developed for ML estimates, where centering does not affect the case-wise scores, and I usually use it with centering=TRUE.

Best,

Rudi

Von: Phil Chalmers [mailto:notifications@github.com] Gesendet: Donnerstag, 6. Juni 2019 02:08 An: philchalmers/mirt mirt@noreply.github.com Cc: RDebelak rudolf.debelak@psychologie.uzh.ch; Author author@noreply.github.com Betreff: Re: [philchalmers/mirt] Centering in estfun function (#161)

Thanks. Can centering = FALSE be the default instead? It seems odd to me that centering the derivatives would be a default to this function.

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/philchalmers/mirt/pull/161?email_source=notifications&email_token=AFOZW3O25BKKIAIJUOQRPCDPZBIMRA5CNFSM4HUBVPR2YY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGODXBLPKY#issuecomment-499300267 , or mute the thread https://github.com/notifications/unsubscribe-auth/AFOZW3J4KEWFUBDAKQA6A5LPZBIMRANCNFSM4HUBVPRQ . https://github.com/notifications/beacon/AFOZW3K2UUQVZHY36KHCRY3PZBIMRA5CNFSM4HUBVPR2YY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGODXBLPKY.gif