part 4 feedback

[andreashandel]

The first part is there, but there isn't much in the analysis section. neither methods describing the analysis, nor results. a few tables are referenced but they don't show up.

It seems the project is not yet as far along as it should be, the majority of the analysis is missing. Make sure you got it all complete by the next submission deadline.

Also, the final version should read like a nice paper/report. that also means reasonably good formatting for text/tables/figures and overall layout. that could use a bit more work.

[rachelr989]

Hi Dr. Handel,

For clarification, is this about the manuscript or the analysis code in the code folder? The methods and results from the analysis we performed have not yet been added to the manuscript but will be soon.

Thank you for your feedback, Rachel

[andreashandel]

I focused on the manuscript. If you've done it, please add the relevant parts (and everything else that still needs updating) to the main manuscript is essentially complete for the next deadline.

[rachelr989]

Thanks for the response. We have done a logistic regression with stepwise selection, a PCA, an RF tree (which needs more tuning), and a LASSO regression (which also needs more tuning). We are having trouble with very multicollinear predictors. We were thinking of doing a Ridge regression as well to help with this problem, but I wanted to ask if this would be repetitive when we already have the LASSO. Should we focus on tuning our existing models more or is there an ideal model that handles multicollinearity very well?

Thank you, Rachel

[andreashandel]

RF and LASSO are some of the best for collinearity. I doubt ridge regression would help much. You can try of course. Otherwise you might want to consider some 'feature engineering' (removing/merging/recoding/etc.) predictors to reduce collinearity.

[rachelr989]

Thank you for the advice! We will look into feature engineering.