Hi,
we have the ANOVA and t tests available for parametric statistical significance hypothesis testing.
are there plans to add any non parametric tests such as chi-squared, fisher exact test, mcnemars test, etc.
simon.
Hi,
we have the ANOVA and t tests available for parametric statistical significance hypothesis testing.
are there plans to add any non parametric tests such as chi-squared, fisher exact test, mcnemars test, etc.
simon.
Ahh, found chi-squared and Fisher exact test, the Crosstab node is my friend.
I cannot find Mcnemars test though.
Simon.
Hi Simon,
So good you support yourself :P
Regarding McNemars, how important is this? I've only just now read the wiki article on it. Once I hear back, I will log this in our tracking system.
Regards,
Aaron
Hi Aaron,
I believe it is important as it has a significant difference and advantage to the chi squared test.
You must have data on the same subject in a before and after situation, I.e. The data is paired (same subject was treated to condition A and then that same subject to condition B).
it effectively means you can achieve statistical significance more easily under these conditions where you have paired data and apply the McNemars test. You could apply the chi squared test but it may not show significance as it fails to consider paired data.
its the non parametric data equivalent of a Paired t-test vs an Independent t-test.
as there are no non-parametric tests in knime that can handle paired data it would add useful functionality. Hopefully it could be incorporated into the cross tab node or would be an additional modified version of it, it is simply the calculation is different but on the same type of contingency table.
Hope that helps
simon.
Perfect Simon, thanks for the more detailed explanation. I've added this to our system and flagged this thread.
Cheers,
Aaron