mrVip: Calculates and helps interpret variable importance for mrIML...

In nfj1380/mrIML: Multi Response Interpretable Machine Learning

Calculates and helps interpret variable importance for mrIML models.

Description

Calculates and helps interpret variable importance for mrIML models.

Usage

mrvip(
  yhats = NULL,
  mrBootstrap_obj = NULL,
  X = X,
  X1 = NULL,
  Y = Y,
  mode = NULL,
  threshold = 0.1,
  global_top_var = 10,
  local_top_var = 5,
  taxa = NULL,
  ModelPerf = ModelPerf
)

Arguments

yhats	A list of model predictions.
mrBootstrap_obj	The object containing model bootstrapping results.
X	The predictor data.
X1	A dataframe extra predictor set used in each model. For the MrIML Joint species distribution model (JSDM) this is just a copy of the response data.
Y	The response data.
mode	character'classification' or 'regression' i.e., is the generative model a regression or classification?
threshold	The threshold for model performance (AUC) below which variables are filtered (default: 0.1).
global_top_var	The number of top global variables to display (default: 10).
local_top_var	The number of top local variables for each response to display (default: 5).
ModelPerf	A list containing model performance metrics for each response variable.

Value

A list containing variable importance data and a combined plot.

Examples

## Not run: 
# Example usage:
#set up analysis
Y <- dplyr::select(Bird.parasites, -scale.prop.zos)%>% 
dplyr::select(sort(names(.)))#response variables eg. SNPs, pathogens, species....
X <- dplyr::select(Bird.parasites, scale.prop.zos) # feature set
X1 <- Y %>%
dplyr::select(sort(names(.)))
model_rf <- 
rand_forest(trees = 100, mode = "classification", mtry = tune(), min_n = tune()) %>% #100 trees are set for brevity. Aim to start with 1000
set_engine("randomForest")
yhats_rf <- mrIMLpredicts(X=X, Y=Y,
X1=X1,'Model=model_rf , 
balance_data='no',mode='classification',
tune_grid_size=5,seed = sample.int(1e8, 1),'morans=F,
prop=0.7, k=5, racing=T) #
ModelPerf <- mrIMLperformance(yhats_rf, Model=model_rf, Y=Y, mode='classification')

bs_analysis <- mrBootstrap(yhats=yhats_rf,Y=Y, num_bootstrap = 5)
bs_impVI <- mrvip(yhats=yhats_rf, mrBootstrap_obj=bs_analysis, ModelPerf=ModelPerf, 
threshold=0.8,  X=X, Y=Y, X1=X1, global_top_var=10,
local_top_var=5)
bs_impVI[[1]] #importance data bootstrap
bs_impVI[[2]] #importance data point estimate
bs_impVI[[3]] #importance plot
bs_impVI[[4]] #PCA
bs_impVI[[5]] #outliers
bs_impVI[[6]] #pc loadings
bs_impVI[[7]] #PC scores

## End(Not run)

nfj1380/mrIML documentation built on May 17, 2024, 7:41 a.m.