| Title: | Calculate Variable Importance with Knock Off Variables |
| Version: | 1.0 |
| Description: | The variable importance is calculated using knock off variables. Then output can be provided in numerical and graphical form. Meredith L Wallace (2023) <doi:10.1186/s12874-023-01965-x>. |
| License: | MIT + file LICENSE |
| Encoding: | UTF-8 |
| RoxygenNote: | 7.2.3 |
| Imports: | caret, ggplot2, ranger, knockoff, ROCR |
| Suggests: | knitr, rmarkdown, testthat |
| VignetteBuilder: | knitr |
| NeedsCompilation: | no |
| Packaged: | 2024-02-20 16:13:09 UTC; wheelerbj2 |
| Author: | Meredith Wallace 
[aut, cre] |
| Maintainer: | Meredith Wallace <lotzmj@upmc.edu> |
| Repository: | CRAN |
| Date/Publication: | 2024-02-21 20:40:06 UTC |
calc_vimps
Description
Calculate the variable importance of the domains for a given dataset
Usage
calc_vimps(
dat,
dep_var,
doms,
calc_ko = TRUE,
calc_dom = FALSE,
num_folds = 10,
num_kos = 100,
model_all = normal_model,
model_subset = one_tree_model,
mtry = NULL,
min.node.size = NULL,
iterations = 500,
ko_path = NULL,
results_path = NULL,
output_file_ko = NULL,
output_file_dom = NULL
)
Arguments
dat |
A dataframe of data |
dep_var |
The dependent variable in the dat |
doms |
A dataframe of the variables in dat and the domain they belong to |
calc_ko |
True/False to calculate the knock_off importance |
calc_dom |
True/False to calculate the domain importance |
num_folds |
The number of folds to use while calculating the classification threshold for predictions |
num_kos |
The number of sets of knock off variables to create |
model_all |
The model to use in full ensemble mode in calculations |
model_subset |
The model to use sigularly for building ensembles from |
mtry |
The mtry value to use in the random forests |
min.node.size |
The min.node.size value to use in the random forests |
iterations |
Number of trees to build while calculating variable importance |
ko_path |
Where to store the knock off variable sets |
results_path |
Where to store the intermediary results for calculating variable importance |
output_file_ko |
Where to store the results of the knock off variable importance |
output_file_dom |
Where to store the results of the domain variable importance |
Value
List with 1) Threshold for binary class labeling 2) Model metrics using all variables 3) Model metrics using knock-off variables 4) Variable importance with knock-offs
Examples
calc_vimps(
data.frame(
X1=c(2,8,3,9,1,4,3,8,0,9,2,8,3,9,1,4,3,8,0,9),
X2=c(7,2,5,0,9,1,8,8,3,9,7,2,5,0,9,1,8,8,3,9),
Y=c(0,0,0,0,0,1,1,1,1,1,0,0,0,0,0,1,1,1,1,1)),
"Y",
data.frame(domain=c('X1','X2'),
variable=c('X1','X2')),
num_folds=2,
num_kos=1,
iterations=50)
graph_results
Description
Graph the variable importance results from calc_vimps
Usage
graph_results(results, object)
Arguments
results |
The results from calc_vimps |
object |
Which object from results to use for graphing results |
Value
No return value