Computes local feature attributions using the lime package
for quantile predictions from a fitted SPQR or SPQRX model.
Explanations are generated on standardized covariates using
the stored normalization parameters.
Usage
eval.explain.lime(
model,
x_training,
x_explain,
tau = 0.5,
n_permutations = 5000,
original_output = FALSE
)Arguments
- model
A fitted model object returned by
fit_spqrx().- x_training
Matrix or data frame of training covariates.
- x_explain
Matrix or data frame of observations to explain.
- tau
Numeric quantile level in (0,1) for quantile prediction.
- n_permutations
Integer. Number of permutations used by
lime::explain().- original_output
Logical. If TRUE, returns the full LIME explanation object. If FALSE (default), returns a reshaped feature-weight table.