EJ
About
 Username
 EJ
 Joined
 Visits
 357
 Last Active
 Roles
 Member, Administrator, Moderator
 Thanked
 46
Comments

BF01 is just 1/BF10, so the same information is presented

If your conclusion depends crucially on whether you use Cohens's d or Hedge's g, this suggests that considerable caution is in order. But JASP provides both (I just saw we had a fix in our Hedges g  a new JASP version comes out end up this month a…

Posted on GitHub and being worked on / implemented, I believe E.J.

Looks good. I'd be consistent and use "Bayes factors" instead of "Bayes Factors". E.J.

I'll also ask that this will be made clear in the output or the help file. E.J.

I think you've since reported this on our GitHub page and we're implementing it... E.J.

Also, when in doubt, you can compare output from R to that of JASP. Cheers, E.J.

There's a 2009 paper in Psychonomic Bulletin & Review by Rouder & Morey: @ARTICLE{RouderEtAl2009Ttest, AUTHOR = {Rouder, J. N. and Speckman, P. L. and Sun, D. and Morey, R. D. and Iverson, G.}, TITLE = {{B}ayesian $t$ Tests for Accepting and…

Hi Joe, population effect size delta = population mu / population sd. So it's the population equivalent of Cohen's d (which is a summary based on the sample, not an inference) Cheers, E.J.

wrt 2: "likelihood" usually refers to something proportional to p(ytheta); so it has a technical connotation that does not match this particular context. wrt 3: Correct, with estimation there isn't a pair of hypotheses  there is only the model th…

Looks very nice! I love the idea of the annotations in general. Will pass this on to the lab as a recommended way to include explanations in tutorial papers. A few suggestions for improvement: I guess you could say that the posterior under H1 is "no…

I'd like to add that, although it is customary to center the prior distribution around the testvalue, this is by no means required. In JASP, the ttests, binomial, multinomial (next release), and AB test (next release) all allow the location of the…

Hi Chris, That's exactly right. This is called the SavageDickey density ratio, and it is a convenient way to "see" a Bayes factor for H0 vs H1 without needing to compute the prior predictive adequacy for H0 and H1 separately. See for instance http:…

yes yes Not sure; what I see from the table is that "Size" is the only factor for which the data offer support (and they do so in compelling fashion) With many factors and interactions, the inclusion ("analysis of effects") approach becomes increasi…

If you are interested in the interaction, you can also compare the 2main effect model against the full model. With few models you don't always need the inclusion BF. Cheers, E.J.

Dear August, The ANOVA as borrowed from the BayesFactor package is really a linear mixed model. But in general, the same assumptions apply as they do for the frequentist version. Yes, the Bayesian approach generally allows all sorts of more flexible…

What needs to be normal is the distribution of variances across subjects, right? I don't see why that wouldn't be (approximately) normal. In your RM ANOVA, I am not sure how your ultimate test involves the variance across the x,y,z. The interaction …

Hi Richard, So far, as you indicate, JASP has focused mostly on the *evidence*, that is, the relative predictive performance for two competing hypotheses (aka the Bayes factor). The reason that we have not put in prior for the hypotheses is that the…

I have attended the relevant JASP team members to this forum entry...

I've passed this on...

Yes. Depends how principled you are about the principle of marginality, I guess :) E.J.

Ah I see. Well but then you can simple compute the sample variance per subject and ttest this between the groups? Of course this ignores the uncertainty about the sample variance, but that would be the approach that people are often using currently…

I'll attend him to your post

This particular analysis includes a numerical routine. This means that unless you take an infinite number of samples (which takes an infinite amount of time) the results will change a little bit from one run to the next. The extent to which the resu…

Yes, it's an issue of numerical precision. Apparently the posterior mass of the models *excluding* that main effect sums to epsilon, with epsilon so close to zero that the number cannot be represented on your computer. E.J.

For an intuitive idea of when a Bayes factor is compelling see https://www.bayesianspectacles.org/letspokeapizzaanewcartoontoexplainthestrengthofevidenceinabayesfactor/ Cheers, E.J.

Hi perdavidson, Sorry for the tardy response. Going by the main output table, the null seems to get most support from the data. The "onlyEmo" model and "onlyVillkor" models also get some support, but a factor of 4 to 5 less. All the other models s…

Sorry for the tardy response. We are very close to a new version (a week from now) and I hope this solves your problem. But, in general, for bug reports and feature requests we hope you can use our GitHub page  this way the problem is accessible t…

Hmm this is interesting, I'll pass this on. Cheers, E.J.

Hi Adam, No I don't think that this is valid in general, although I would not be surprised if it is a fairly good approximation under many circumstances. This would be an interesting topic for a statistical study. Well you'd need an interval or SE…