Statistical significance of correlation

Dear KNIMErs,

What do you think about adding the r² and the p-value to the output of the linear correlation node? See here for details (source in German):

It'd be most useful, I assure you! :-)


Dear Ergonomist,

thanks for the pointer. We can add it (though I'm afraid that our project plan is too dense that we can allocate additional resources for 2.0).

Currently similar statistics are available in the node view of the learner node (there are two values for squared error and total squared error). I can imagine that we add an additional output table, which contains these statistics in a single row for further processing. (In 1.3.x the node has a single output with the linear regression model.)


Sounds good! I don't really mind waiting for 2.1 with that. ;-)