"Cannot plot posterior--possibly too concentrated near 0" does not display theta and delta 95% CI
Hello!
This is the first time coming across an error where it is not displaying the pop. effect size and its 95% CI.
The error displayed: "Cannot plot the posterior - possibly too concentrated near 0."
- I've ran a bayesian one samples t-test against 0. The alternative hypothesis is theta 0.
- Prior @ Default.
- BF+0 = 3.83 x 10^7
I need to report a delta value and its 95% CI. What are some recommendations on ways I can find these values?
If it helps, descriptives:
- Mean = 55.560
- SD = 14.260
- SE = 3.682
- COV = 0.257
- 95% Credible interval = [47.663, 63.457]
Much thanks in advance!
Comments
theta greater than 0*
Are you sure you want to test against a value of zero? The plot fails because the result is so extreme, but the table still ought to provide the information, right?
EJ
Hi EJ,
Yes! I am testing for a pop-out effect (for multiple conditions). If searching for a target is really easy and it seems to "pop out" amongst distractors, I expect the reaction time to be very low when the number of distractors is low and high. My dependent variable is a search slope--if pop-out occurs, the search slope should be very close to zero. If search is pretty challenging with larger amounts of distractors, I'd expect large search slopes.
Unfortunately, the table does not include the population effect size and 95% Credible intervals for one of the conditions (I've been referencing the values that appear at the top of the plots).
Does it work when you do a two-sided test? This looks more like a bug that would be useful to take to our GitHub page (https://jasp-stats.org/2018/03/29/request-feature-report-bug-jasp/)
EJ
Unfortunately, it does not. Bug report--will do.
Thanks, this is now taken to https://github.com/jasp-stats/jasp-issues/issues/2603
EJ
Hi, I am new here, and scanning through the posts.
If it helps, I have noticed the same error with a few of my datasets.
However, just to see if there was a difference, I used the Bayesian JASP module in JAMOVI - and it provided the missing table. I significantly prefer JASP; however, this might work in the meantime while the bug is fixed. (And perhaps find the bug if comparing the two modules).
Jon
That is extra strange, and provided additional motivation to fix this :-). I've attended the team to the issue.
EJ