Avatar

Howdy, Stranger!

It looks like you're new here. If you want to get involved, click one of these buttons!

Supported by

EJ

About

Username
EJ
Joined
Visits
274
Last Active
Roles
Member, Administrator, Moderator
Thanked
38

Comments

  • Hi Dario, "calculate how much the old version of the test is correlated to the new one" Since the old and new test share a lot of items, the correlation has to be high. To compute the correlation, ideally you have the same people answer all the …
  • Hi Dario, Well, first off, data from a Likert scale can never be normal, because the scale is discrete. But sum scores across several items (or average scores across participants, or averaged sum scores across participants) can be approximately n…
  • Hi Shun, Yes you can; it is under the plotting options in the ANOVA menu Cheers, E.J.
  • Hmmm. There are at least two solutions to this problem. The first is simple: just eyeball the posterior distributions of the beta's. Although informative, this is of course not a formal test. The second solution would be to compare the models with t…
  • We've made a lot of progress since then, so I think we are in a good position to address this issue now.
  • Ah, so we need to be able to save the factor scores as a separate column. I think that this should in principle be doable; if you GitHub this then I can ask those responsible for the factor analysis code to do this (maybe it is already done, at leas…
  • I recall discussing this before, at least when the combination rule is of a simple form (e.g., a weighted average over the columns). Assuming that this is what you mean, I'll ask some team members for advice.
  • Hi Dirk, Exp 1 provided some evidence against the interaction; consequently, in the model that includes the interaction, the corresponding posterior distribution will have more mass near zero than the prior did. In other words, the interaction --…
  • Hi Mark, When you say "extract a single factor", do you refer to factor analysis or just to averaging? Cheers, E.J.
  • That's right, but you can prompt us to do so by adding a feature request on our GitHub page (for details see https://jasp-stats.org/2018/03/29/request-feature-report-bug-jasp/)! Cheers, E.J.
  • Hi Sunny, This is because in ANOVA, the BF is obtained by numerical methods. So the result is an approximation (the % error tells you how good that approximation is). If you want to decrease the % error, you can go to advanced options and increas…
  • Dear rohanp16, This does look like a bug. I'll report it for you on our GitHub page (https://jasp-stats.org/2018/03/29/request-feature-report-bug-jasp/) Cheers, E.J.
  • Hi Dirk, There are two ways to do this. First, there is the Verhagen & Wagenmakers method, where you "simply" use the posterior from the first experiment as a prior for the second experiment. Unfortunately, the updating and specification proc…
  • Hmm I noticed I basically re-entered my earlier answer. Well, goes to show I didn't change my mind about this. E.J.
  • Hi NvH, Yes, this has been argued, mostly by subjective Bayesians such as Lindley. The Bayesian "correction" for multiplicity is in the prior model probability. If you are testing 10 effects, do you really believe that every single test is plausi…
  • Hi Franziska, That is the same thing, isn't it? The shape of the boxplot is a little different, but the information appears to be identical (?) Cheers, E.J.
  • Hi Ondrej, Great question. Ideally, you'd integrate the ANOVA structure with the model underneath. When that is difficult some two-step method could be used. Here's a paper that may be relevant: https://www.collabra.org/article/10.1525/collabra.7…
  • Good question! The .707 from the default test is really 1/2 * sqrt(2). What you can do to check is use the informed test with "0.7071068" instead -- the result should be closer. Also, we have a better routine for the informed t-test now (much faster…
  • Yes, this should be resolved. You can set the preference for the number of decimal points in Preferences. See attached screenshot. Cheers, E.J.
  • Hi Philip, The %error gives an indication of how accurately the numerical methods have estimated the BF. I think that it's the standard error expressed as a proportion of the BF, but the BayesFactor package should explain what it is exactly. When…
  • Dear Michif, I'll ask Johnny to clarify this. We probably call an existing R package. Cheers, E.J.
  • * If you set BF to "BF01" instead of "BF10", then you'll see how many more times the best model predicted the data (numbers > 1 are usually easier to interpret). * The most popular way to test an interaction is to compare A+B to A+B+A*B. The eas…
  • Hi Patrick Yes, the copy-paste functionality can be improved. The default copy-paste for tables copies the html code, but I don't think that this is ideal. We also have "copy as" and then offer LaTeX code, but I hope we can add ASCII in the futur…
  • Yeah this should not happen. It may be an issue that we've already resolved. Advice: 1. Please post the issue on our GitHub page, so the team can take a look (see https://jasp-stats.org/2018/03/29/request-feature-report-bug-jasp/) 2. A new version…
  • Hi Herry, * It may help if you select "compare to best model" * Let's break this down. Model A beats the null model by a tremendous amount. Model B actually does worse than the null model. Model A + B outpredicts the null model by a lot, but n…
  • So the prior odds have been adjusted according to the Westfall formula. If you multiply this by the uncorrected "regular" BF you will get the posterior odds, which you can interpret as a "corrected" BF.
  • Yes, that could in general be the case, although your specific example looks like a ratio scale. Of course you may be interested in an underlying psychological process that has a monotonic relation to the number of words; in this case an analysis as…
  • Hi DJ, Unfortunately, this cannot be done yet. We are working on a comprehensive solution for editing all plots, but that will take a while to get done (we have to re-program all of our figures, and that's just the start). Cheers, E.J.
  • Hi Mila, If you have accuracy data, then for a single individual you can use the binomial, for which the CI respects the bound. However, I assume that you have multiple individuals, and you analyze the proportion correct across subjects. You can …
  • yeah, taking that interrelationship into account would be best but I am not sure it is worth the trouble E.J.