Correlation filter:Which variable is choosen to survive if there are 2 varibales only correlating with each other?

KNIME Analytics Platform
1

Hi,

 

I have a question regarding the correlation filter. Which variable is choosen to survive if there are 2 varibales only correlating with each other? 

 

Thank you!

Andra

2

If you open the dialog of the Correlation Filter node you can check yourself, which columns are included based on the threshold you set.

3

Hi,

 

thanks for your reply! However, I still do not understand the logic behind it. If I have two Variables correlating with a value of 0.8 and my threshold is 0.7, one of those will be disregarded. But which one and why?

Thanks

Andra

4

It's determined based on the order of the columns. There is no deep theory behind it; it's just greedy. If you reverse the column order the output will be different.

We have a feature request open to allow the user to specify some priority constaints.

Hope this helps,
  Bernd