Logistic Regression

KNIME Analytics Platform
1

Hello,

 

I would like to do a logit model in KNIME and use Logistic Regression nodes. The problem is that I don't understand the "Coefficients and Statistics" output of Logistic Regression Learner. My target is a nominal variable with 2 values: "yes" and "no". It has no missing values. In the "Coefficients and Statistics" I see Logit coefficients for "no" and 0.

1. Where did 0 appear from?

2. Why does the model give the coefficients with regards to "no" and not "yes"?

I attach the print screen of Logistic Regression Learner "Coefficients and Statistics" output.

 

Another thing is that in Scorer node I can only see the accuracy of the whole model but where can I get the statistical significance of the model (e.g. Likelihood ratio test)?

 

Thank you in advance,

F

2

I can try to anwser the bonus question: the Scorer node has a second output with the Accuracy statistics including TruePositives, ..., FalseNegatives, Recall, Precision, Sensitivity, Specifity, F-measure and Accuracy -- if this helps at all.

3

Hi F:

Have you tried to used a Clasiffication Tree or an Artificial Network?  Month ago I tried to do a Logistic Regression, but the report that gave KNIME it was insufficient to understand the model. A good alternative is Tanagra. That open source software has an excellent logistic regression.

Bye

Gabriel

 

4

Hi Gabriel Cornejo,

Thank you for your advice!

Should it be in Regression section of Tanagra? I cannot find Logistic Regression there.

F

5

I realize I am very late to this party but I had a similar question and saw this post. You can set the reference category in the Log Reg Learner. Since this is the reference category, if you set it to "yes" then "no" will appear under logit in the coefficients output table.