predict.mvrm {BNSP}R Documentation

Model predictions

Description

Provides predictions and posterior credible/prediction intervals for given feature vectors.

Usage

## S3 method for class 'mvrm'
predict(object, newdata, interval = c("none", "credible", "prediction"), 
                       level = 0.95, nSamples = 100, ...)

Arguments

object

an object of class "mvrm", usually a result of a call to mvrm.

newdata

data frame of feature vectors to obtain predictions for. If newdata is missing, the function will use the feature vectors in the data frame used to fit the mvrm object.

interval

type of interval calculation.

level

tolerance level.

nSamples

number of samples to obtain from the posterior predictive distribution (for each sweep of the MCMC). Only relevant for "prediction intervals".

...

other arguments.

Details

The function returns predictions of new responses or the means of the responses for given feature vectors. Predictions for new responses or the means of new responses are the same. However, the two differ in the associated level of uncertainty: response predictions are associated with wider (prediction) intervals than mean response predictions. To obtain prediction intervals (for new responses) the function samples from the normal distributions with means and variances as sampled during the MCMC run.

Value

Predictions for given covariate/feature vectors.

Author(s)

Georgios Papageorgiou gpapageo@gmail.com

See Also

mvrm

Examples

#see \code{mvrm} example

[Package BNSP version 2.1.6 Index]