Hw #4

Hw #4

by Grishma -
Number of replies: 3

For question 13, I am not sure what it is asking me to do. How would I "perform separate analysis for each predictor to determine whether it is correlated with the class label" ?

In reply to Grishma

Re: Hw #4

by John Kornak -

You do not need to actually perform this analysis in Modeler. The question is just looking for a general descriptive answer.

When screening predictors in an analysis you fit a separate model for each predictor in turn to try and predict the outcome. You then select only those predictors that do better than some level (e.g. correlation > 0.8). You would only consider predictors further for larger scale modeling (i.e. multiple predictors) if they survived this initial level of screening.

I hope that helps?

John

In reply to John Kornak

Re: Hw #4

by Grishma -

Thank you for your responses Daniel and John. That was precisely what I was wondering about i.e.  whether it required a general descriptive answer or an actual analysis.