Enhancement: Jensen-Shannon & Kolmogorov-Smirnov Tests

jasp-stats / jasp-issues

This repository is solely meant for reporting of bugs, feature requests and other issues in JASP.

59 stars 29 forks source link

Enhancement: Jensen-Shannon & Kolmogorov-Smirnov Tests #210

Open TarandeepKang opened 6 years ago

TarandeepKang commented 6 years ago

Jensen-Shannon & Kolmogorov-Smirnov Tests

Enhancement: Addition of the ability to perform JS & KS test
Purpose: Standard and widely used tests (at least in my field of animal cognition) to test wether a pair of curves come from the same distribution. This would mean all necessary features (except flexibility in graphing) are implemented and I can complete my transition to JASP!

It was a pleasure to meet EJ at the BPS/EPS Reproducibility event at the Royal Society a few months ago. Great work JASP Team!

AlexanderLyNL commented 6 years ago

Dear TarandeepKang,

Thanks for the suggestion. We still have to think about where we would place these tests. Please let us know if you have any suggestions.

Cheers, Alexander

TarandeepKang commented 6 years ago

Hi Alexander, yes I have an idea. At the moment every form of nonparametric test (ANOVA, ANCOVA etc) is included as an option within the sub menu for that analysis procedure. Would it be conceivable able to separate out and reorganise all the nonparametric tests into a separate menu item and then include my suggested tests along with your existing nonparametric functionality?

Best,

Tarandeep

EJWagenmakers commented 6 years ago

Although possible, this is too much of a reorganization for my taste. I prefer to keep the tests where people may use them. We could potentially add a menu item "Assumption Checks" below "Descriptive Statistics" and above "Reliability Analysis"; this is where we could place tests for normality, equal variance, whether samples come from the same distribution, etc.

TarandeepKang commented 6 years ago

Sounds good EJ!

tomtomme commented 10 months ago

KS-Test now exists in the distribution module. Other distribution tests alongside related analysis. However as far as I can tell Jensen-Shannon divergence is missing still. The alternative KL-divergence however is available Bayesian Unidimensional Reliability.

@TarandeepKang Does this suffice for your needs? If yes, feel free to close this issue. If not, we may reevaluate where the JS divergence may fit.

TarandeepKang commented 9 months ago

I think the current implementations look good, and as far as I'm concerned we can consider closing this?

tomtomme commented 9 months ago

@TarandeepKang Its your issue :) Close it, if you feel like it or leave it open if JS is important to you