toDos (in progress)

[juliuspfadt]

Bugs:

• [x] Ensure the percentage in the uncertainty interval can be changed.
• [x] Ensure changing the percentage of the ‘uncertainty interval’ does not cause errors.
• [x] Ensure equivalent hypotheses under the standard and manual test result in the same posterior probabilities.
• [x] Ensure the standard hypothesis test gives the same posterior probabilities as the manual test for equivalent hypotheses in independent samples t-test
• [ ] Address the issue with empty coefficients table in independent t-test for specific data (e.g., sesame data with variables “age” and “sex”): Problem is definitely in BFpack, which does not export $estimates for that specific test. @jomulder
• [x] Ensure the standard hypothesis test gives the same posterior probabilities as the manual test for equivalent hypotheses for paired samples t-test
• [x] Fix the issue with strange posterior probabilities when the test value is the same as the mean in one sample t-test: Could not replicate anymore
• [x] Ensure the standard hypothesis test gives the same posterior probabilities as the manual test for equivalent hypotheses for one sample t-test
• [ ] ANCOVA stuck in loop? Likely caused by the interaction which, as an R source, restarts the analysis constantly. We need a qml element for interactions for two-way, three-way etc (dynamically ideally).

Features:

• [x] correlation:
  • [x] grouping variable: I added a footnote to the first output table that indicates the levels of the groups.
  • [x] covariates
• [ ] Implement a “standardize coefficients” option similar to the bain module: I dont think this is as easy as just using scale on the data before the analysis. That is at least not what Bain does, the option is only available for ANOVA and regression and the hypotheses are then evaluated on the standardized coefficients.
• [ ] implement cmd+Enter for running the manual hypotheses
• [ ] Provide options for users to select which interactions should be included.
• [ ] Add support for multiple dependent variables for multivariate regression (???)
• [ ] Handle interactions similar to Linear Regression.
• [ ] Add support for multiple dependent variables for MANOVA. (I feel like that is already there)
• [ ] mvt-test: Add support for the one-sample version, independent samples version, and paired samples version.
• [ ] mvt-test: Align the test values in the “Test value” box to the right.
• [ ] Add support for the meta-analytic model in BFpack.

[juliuspfadt]

So the thing with the different results for equivalent standard and manual hypotheses is due to the BF.type argument that is BF.type=1 by default in JASP, meaning the fractional BF is the default. This actually leads to the differences. The default in BFpack however is BF.type = 2, meaning the adjusted fractional BF. so now I changed the default in JASP as well, et voila it works as expected.

[jomulder]

So the thing with the different results for equivalent standard and manual hypotheses is due to the BF.type argument that is BF.type=1 by default in JASP, meaning the fractional BF is the default. This actually leads to the differences. The default in BFpack however is BF.type = 2, meaning the adjusted fractional BF. so now I changed the default in JASP as well, et voila it works as expected.

Ah that's lame. Sorry about that!