politenessProjection {politeness}

Politeness projection

Description

Training and projecting a regression model using politeness features.

Usage

politenessProjection(
  df_polite_train,
  covar = NULL,
  df_polite_test = NULL,
  classifier = c("glmnet", "mnir"),
  cv_folds = NULL,
  ...
)

Arguments

df_polite_train

a data.frame with politeness features as outputed by politeness used to train model.

covar

a vector of politeness labels, or other covariate.

df_polite_test

optional data.frame with politeness features as outputed by politeness used for out-of-sample fitting. Must have same feature set as polite_train (most easily achieved by setting dropblank=FALSE in both calls to politeness).

classifier

name of classification algorithm. Defaults to "glmnet" (see glmnet) but "mnir" (see mnlm) is also available.

cv_folds

Number of outer folds for projection of training data. Default is NULL (i.e. no nested cross-validation). However, positive values are highly recommended (e.g. 10) for in-sample accuracy estimation.

...

additional parameters to be passed to the classification algorithm.

Details

List:

• train_proj projection of politeness model within training set.

• test_proj projection of politeness model onto test set (i.e. out-of-sample).

• train_coef coefficients from the trained model.

Value

List of df_polite_train and df_polite_test with projection. See details.

Examples

data("phone_offers")
data("bowl_offers")

polite.data<-politeness(phone_offers$message, parser="none",drop_blank=FALSE)

polite.holdout<-politeness(bowl_offers$message, parser="none",drop_blank=FALSE)

project<-politenessProjection(polite.data,
                              phone_offers$condition,
                              polite.holdout)

# Difference in average politeness across conditions in the new sample.

mean(project$test_proj[bowl_offers$condition==1])
mean(project$test_proj[bowl_offers$condition==0])