Begüm D. Topçuoğlu

@zmml Would you like to do a PR if you made any changes on this? I've submitted the first release and can start working on these issues. Thank you for...

```get_results(dataset, models, split_number, outcome="dx")``` errors out when `outcome` is defined by user. ```pipeline(dataset, models, split_number, outcome="dx")``` works. So it must be an issue when the argument is being passed from...

Define outcome variable and permutation logical as arguments that can be passed in the command line. We can now run from command line with: `Rscript code/learning/main.R 1 "L2_Logistic_Regression" "dx" 0`

Previous changes regarding setting outcome and perm on the command line work when the user defines the outcome (e.g. "dx") as an argument. However, if they leave that argument empty...

With these changes, it looks like we Fixed #6 and #7. Next step is checking if permutation works as we want in #8.

`permutation_importance` function doesn't work. Error: ``` Error in -sym(first_outcome) : invalid argument to unary operator Calls: get_results ... -> vars_select_eval -> map_if -> map -> .f ``` Caught the bug...

> Nooo I was worried this wouldn't work but I wasn't doing permutation importance so I didn't catch it. I can look into another option if you don't know of...

> But don't we want to have dx not hard-coded? No I know, I have it as: ``` non_correlated_otus % select(-correlated_otus) %>% select(-sym(outcome)) %>% colnames() ``` Instead of what it...

Our previous attempt was unsuccessful - must be a bug with tidyverse. I made a new change: ``` non_correlated_otus % select(-correlated_otus) non_correlated_otus[,outcome]

#8 We now made permutation importance optional but the data structure to run permutation is still hardcoded. We need to come back to that and fix it. I'll now check...