No effect of changing the prior in t-tests

[lindeloev]

Nothing changes if you change the prior. Not the BF, not the prior/posterior plot, etc.

To reproduce:

1. Do a Bayesian t-test (one-sample or two-sample or paired), e.g., on the sleep data.
2. (not necessary) tick off everything to see that nothing is affected.
3. Change the prior to something outrageous. E.g., a normal with location 9. Nothing changes.

This works as expected in JASP. Changing priors in ANOVA seems to work OK.

[jonathon-love]

hi, have you checked the 'informative' radio button? seems to work here:

https://youtu.be/TZwlCqGQzAk

with thanks

jonathon

[lindeloev]

Oh yes, you are right that that's the issue! There may be some usability here. I would expect it to follow the JASP behavior: (1) Have all informative grayed out if Default is and likewise gray out the text field in Default when an informative is selected, and (2) Use the latest prior that one interacted with (which would occur automatically if one did (1), I think). Best, Jonas

…

On Thu, May 16, 2019 at 2:49 AM Jonathon Love ***@***.***> wrote: hi, have you checked the 'informative' radio button? seems to work here: https://youtu.be/TZwlCqGQzAk with thanks jonathon — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#2?email_source=notifications&email_token=AAK7NOKDENZFEU66BTVDMELPVSVRXA5CNFSM4HNCENIKYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGODVQKYWA#issuecomment-492874840>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AAK7NOIQSAD57IWD4ZYBDV3PVSVRXANCNFSM4HNCENIA> .

[jonathon-love]

oh yup. i'll take a look in the new little while.

cheers and thanks for raising this.