predict.mixpoissonreg: Predict Method for 'mixpoissonreg' Objects

In mixpoissonreg: Mixed Poisson Regression for Overdispersed Count Data

Description

Function to obtain various predictions based on the fitted mixed Poisson regression models.

Usage

## S3 method for class 'mixpoissonreg'
predict(
  object,
  newdata = NULL,
  type = c("response", "link", "precision", "variance"),
  se.fit = FALSE,
  interval = c("none", "confidence", "prediction"),
  level = 0.95,
  nsim_pred = 100,
  nsim_pred_y = 100,
  ...
)

Arguments

object	object of class "mixpoissonreg" containing results from the fitted model.
newdata	optionally, a data frame in which to look for variables with which to predict. If omitted, the fitted response values will be provided.
type	the type of prediction. The default is the "response" type, which provided the estimated values for the means. The type "link" provides the estimates for the linear predictor. The type "precision" provides estimates for the precision parameters whereas the type "variance" provides estimates for the variances.
se.fit	logical switch indicating if standard errors on the scale of linear predictors should be returned. If TRUE, it only returns the standard deviations of the linear predictors when type = 'link', otherwise returns NA and a warning indicating that the type must be 'link'. When using mixpoissonreg objects, the fit must be done with x = TRUE.
interval	Type of interval calculation for the response variables, 'none', 'confidence' or 'prediction'. If 'confidence', the confidence intervals for the means are returned. If 'prediction', prediction intervals for future response variables are reported. For confidence intervals, the type of the prediction must be 'response' or 'link'. For prediction intervals the type of prediction must be 'response'. For 'confidence' intervals, when using mixpoissonreg objects, the fit must be done with x = TRUE and for predictions intervals, the fit must be done with x = TRUE and w = TRUE.
level	Tolerance/confidence level. The default is set to 0.95.
nsim_pred	number of means and predictions to be generated in each step of the simulation. The default is set to 100.
nsim_pred_y	number of response variables generated for each pair of mean and precision to compute the prediction intervals. The default is set to 100.
...	further arguments passed to or from other methods.

Details

The se.fit argument only returns a non-NA vector for type = 'link', that is, on the scale of the linear predictor for the mean parameter. For the response scale, one can obtain confidence or prediction intervals. It is important to notice that confidence intervals must not be used for future observations as they will underestimate the uncertainty. In this case prediction intervals should be used. Currently, we do not have closed-form expressions for the prediction interval and, therefore, they are obtained by simulation and can be computationally-intensive.

Value

A vector containing the predicted values if se.fit=FALSE, a list with elements fit and se.fit if se.fit=TRUE, and a matrix if interval is set to confidence or prediction.

See Also

fitted.mixpoissonreg, summary.mixpoissonreg, plot.mixpoissonreg, autoplot.mixpoissonreg, coef.mixpoissonreg, vcov.mixpoissonreg, plot.mixpoissonreg

Examples

daysabs_prog <- mixpoissonreg(daysabs ~ prog, data = Attendance)
predict(daysabs_prog)

mixpoissonreg documentation built on March 11, 2021, 1:07 a.m.