Adding two column "Feature selection" to component configuration

I am creating a version of the logistic regression learner with different/additional output. The component is all setup with the original learner embedded within it, all of the correct changes/additions done with other nodes and the correct input and output ports. It also has a Column Selection Configuration node that lets the component configuration set the target cell within the learner node.

Is there any way to add a configuration node that will also add the two-column (Exclude-Include) “Feature selection” for the learning into the component configuration, or will the user always have to go into the component to make that selection?

I’m new to KNIME and especially components, so I appreciate any help that can be provided.

Side question: anyone know a trick in KNIME to find a (right-tail) chi-square p-value if I already have the chi-square test statistic and the DOF. That’s the one output that I can’t provide yet. I assume it could be done with R or Python, but I’m looking for a self-contained way.

Hi @CWrightNU, welcome to the community! Nice to hear you are already building components :slight_smile:

There is the the Column Filter Configuration that provides the twinlist - no need to open up the component each time.

Regarding your side question: have you had a look at the Crosstab Node? In its second output it provides the “Chi-Square (Prop)” column, which should be the p-value.

Kind Regards,
Lukas

2 Likes

Thanks, Lukas! Not sure how I missed that. Maybe I saw “Column” and thought it was the “Column Selection” one that I had already used. It’s working as intended now!

As for the Crosstab Node, I did look at it, but it creates the test statistic from the data. I couldn’t figure out how to trick it into calculating the p-value if I already have the test statistic calculated.

1 Like

This topic was automatically closed 7 days after the last reply. New replies are no longer allowed.