ndlVarimp function

Permutation variable importance for classification using naive discriminative learning.

ndlVarimp uses permutation variable importance for naive discriminative classification models, typically the output of ndlClassify.

ndlVarimp(object, verbose=TRUE)

Arguments

• object: An object of class "ndlClassify" (or one that can be coerced to that class); typically a model object as produced by ndlClassify.
• verbose: A logical (default TRUE) specifying whether the successive predictors being evaluated should be echoed to stdout.

Details

Variable importance is assessed using predictor permutation. Currently, conditional permutation variable importance (as for varimp

for random forests in the party package) is not implemented.

Returns

A list with two numeric vectors:

• concordance: For binary response variables, a named vector specifying for each predictor the index of concordance when that predictor is permuted. For polytomous response variables, NA.
• accuracy: A named vector specifying for each predictor the accuracy of the model with that predictor permuted.

References

R. Harald Baayen (2011). Corpus linguistics and naive discriminative learning. Brazilian journal of applied linguistics, 11, 295-328.

Carolin Strobl, Anne-Laure Boulesteix, Thomas Kneib, Thomas Augustin and Achim Zeileis (2008). Conditional Variable Importance for Random Forests. BMC Bioinformatics, 9, 307.

Author(s)

R. H. Baayen and Antti Arppe

See Also

summary.ndlClassify, plot.ndlClassify, anova.ndlClassify,ndlCuesOutcomes, estimateWeights, cueCoding

Examples

## Not run:

data(dative)
dative <- dative[!is.na(dative$Speaker),-2]
dative.ndl <- ndlClassify(RealizationOfRecipient ~ ., data=dative)
dative.varimp <- ndlVarimp(dative.ndl)

library(lattice)
dotplot(sort(summary(dative.ndl)$statistics$accuracy-dative.varimp$accuracy), 
   xlab="permutation variable importance")
## End(Not run)