predict.RprobitB_fit: Predict choices

In RprobitB: Bayesian Probit Choice Modeling

Predict choices

Description

This function predicts the discrete choice behavior

Usage

## S3 method for class 'RprobitB_fit'
predict(object, data = NULL, overview = TRUE, digits = 2, ...)

Arguments

object	An object of class RprobitB_fit.
data	Either NULL, using the data in object, an object of class RprobitB_data, for example the test part generated by train_test, or a data frame of custom choice characteristics. It must have the same structure as choice_data used in prepare_data. Missing columns or NA values are set to 0.
overview	If TRUE, returns a confusion matrix.
digits	The number of digits of the returned choice probabilities. digits = 2 per default.
...	Ignored.

Details

Predictions are made based on the maximum predicted probability for each choice alternative. See the vignette on choice prediction for a demonstration on how to visualize the model's sensitivity and specificity by means of a receiver operating characteristic (ROC) curve.

Value

Either a table if overview = TRUE or a data frame otherwise.

Examples

data <- simulate_choices(
  form = choice ~ cov, N = 10, T = 10, J = 2, seed = 1
)
data <- train_test(data, test_proportion = 0.5)
model <- fit_model(data$train)

predict(model)
predict(model, overview = FALSE)
predict(model, data = data$test)
predict(
  model,
  data = data.frame("cov_A" = c(1, 1, NA, NA), "cov_B" = c(1, NA, 1, NA)),
  overview = FALSE
)

RprobitB documentation built on May 29, 2024, 7:59 a.m.