Bug: Correlation coefficient is outside CI for one-sided tests

[lindeloev]

JASP 0.9.1, Windows 10, 64-bit.

The CI reacts to the hypothesis but the point estimate does not. If the point estimate has opposite sign of the hypothesis, this is particularly clear since the estimate is outside the CI (see top row with WMI.change):

This is true for both Bayesian correlation pairs and Bayesian correlation matrix. I would expect everything to conform to the hypothesis (model). The plot seems to get it right.

[EJWagenmakers]

"r" is the sample Pearson correlation coefficient, which is a test statistic calculated from the data without knowledge of the hypotheses. But I understand the source of your confusion: one would perhaps expect the posterior median for rho in the same table that shows the credible interval. This is entirely legit, and it is a little weird that we do provide the interval in the table but not the point estimate. @AlexanderLyNL : when you get to this, I think it would be a good idea to show the posterior median under H1 in a table as well (also for the t-test, for delta).

[lindeloev]

Yeah, I would actually expect everything in the Bayesian analyses to be posteriors unless explicitly stated otherwise ("descriptives", N, etc.) Not sure if my expectations are shared by the average user, though.

[EJWagenmakers]

Yes I think the potential confusion between "r" and "rho" will be reduced if we provide a point estimate for rho as well as a CI. E.J.

[tomtomme]

Still valid for jasp 0.19 beta