mrFlashlight: Wrapper to run multi-response model agnostic interpretable machine learning analyses.

Usage

mrFlashlight(
  yhats,
  X,
  X1,
  Y,
  response = "multi",
  index = 1,
  mode = "regression",
  predict_function = NULL
)

Arguments

yhats: A list generated by mrIMLpredicts.
X: A data.frame containing the feature data set.
X1: A data.frame of additional predictors (optional).
Y: A data.frame containing the response variable data set (species, OTUs, SNPs, etc.).
response: character indicating the type of response: 'single' selects one response, 'multi' selects all responses.
index: numeric used when response is 'single' to select which response to create a flashlight object for. The order is the same as in Y.
mode: character indicating the type of model: 'classification' or 'regression'.
predict_function: function specifying a user-defined prediction function (optional).

Details

This function enables users to utilize interpretable machine learning methods to understand their multi-response and single-response models.

Examples

# Single response
fl <- mrFlashlight(yhats, X, Y, response = "single", index = 1, mode = "regression")
#> Error in flashlight(model = yhats[[index]]$mod1_k, label = colnames(Y)[index],     data = cbind(Y[index], X), y = colnames(Y)[index], predict_function = pred_fun,     metrics = metrics): could not find function "flashlight"
plot(light_performance(fl), fill = "orange") + labs(x = element_blank())
#> Error in light_performance(fl): could not find function "light_performance"
plot(light_breakdown(fl, new_obs = cbind(X, Y)[1, ]), by = X, v = Y)
#> Error in light_breakdown(fl, new_obs = cbind(X, Y)[1, ]): could not find function "light_breakdown"
int <- light_interaction(fl, pairwise = TRUE)
#> Error in light_interaction(fl, pairwise = TRUE): could not find function "light_interaction"

# Multiple response
flashlightObj <- mrFlashlight(yhats, X, Y, response = "multi", mode = "regression")
#> Error in eval(expr, envir, enclos): object 'yhats' not found