predictive_error.stanreg {rstanarm}

In-sample or out-of-sample predictive errors

Description

This is a convenience function for computing y - y^{rep} (in-sample, for observed y) or y - \tilde{y} (out-of-sample, for new or held-out y). The method for stanreg objects calls posterior_predict internally, whereas the method for matrices accepts the matrix returned by posterior_predict as input and can be used to avoid multiple calls to posterior_predict.

Usage

## S3 method for class 'stanreg'
predictive_error(
  object,
  newdata = NULL,
  draws = NULL,
  re.form = NULL,
  seed = NULL,
  offset = NULL,
  ...
)

## S3 method for class 'matrix'
predictive_error(object, y, ...)

## S3 method for class 'ppd'
predictive_error(object, y, ...)

Arguments

object

Either a fitted model object returned by one of the rstanarm modeling functions (a stanreg object) or, for the matrix method, a matrix of draws from the posterior predictive distribution returned by posterior_predict.

newdata, draws, seed, offset, re.form

Optional arguments passed to posterior_predict. For binomial models, please see the Note section below if newdata will be specified.

...

Currently ignored.

y

For the matrix method only, a vector of y values the same length as the number of columns in the matrix used as object. The method for stanreg objects takes y directly from the fitted model object.

Value

A draws by nrow(newdata) matrix. If newdata is not specified then it will be draws by nobs(object).

Note

The Note section in posterior_predict about newdata for binomial models also applies for predictive_error, with one important difference. For posterior_predict if the left-hand side of the model formula is cbind(successes, failures) then the particular values of successes and failures in newdata don't matter, only that they add to the desired number of trials. This is not the case for predictive_error. For predictive_error the particular value of successes matters because it is used as y when computing the error.

See Also

posterior_predict to draw from the posterior predictive distribution without computing predictive errors.

Examples

if (.Platform$OS.type != "windows" || .Platform$r_arch != "i386") {
if (!exists("example_model")) example(example_model)
err1 <- predictive_error(example_model, draws = 50)
hist(err1)

# Using newdata with a binomial model
formula(example_model)
nd <- data.frame(
 size = c(10, 20), 
 incidence = c(5, 10), 
 period = factor(c(1,2)), 
 herd = c(1, 15)
)
err2 <- predictive_error(example_model, newdata = nd, draws = 10, seed = 1234)

# stanreg vs matrix methods
fit <- stan_glm(mpg ~ wt, data = mtcars, iter = 300)
preds <- posterior_predict(fit, seed = 123)
all.equal(
  predictive_error(fit, seed = 123),
  predictive_error(preds, y = fit$y)
)
}