predict.censReg: Predict Values

In USGS-R/smwrQW: Tools for censored data analysis

Predict Values

Description

Predicts values from a censored regression object: method for "censReg" object.

Usage

## S3 method for class 'censReg'
predict(object, newdata, type = c("response", "link",
  "mean"), se.fit = c("none", "estimate", "prediction"),
  interval = c("none", "confidence", "prediction"), level = 0.95,
  na.action = na.pass, ...)

Arguments

object	the output from censReg.
newdata	an optional data frame in for predictions. If omitted, the observations from object are used.
type	the type of predicted value. To get predictions on the original scale of the response, use "response." To get predictions on the transformed scale of the response, use "link." To get the mean predictions for log-normal data, use "mean."
se.fit	the type of standard errors of the predicted value to include in the result.
interval	type type of interval calculation. See Details.
level	the confidence or prediction interval.
na.action	function determining what should be done with missing values in newdata. The default is to predict NA.
...	further arguments passed to or from other methods.

Details

Some combinations of type and se.fit and interval are not possible, depending on the distribution of the model. If type is "link" or the distribution is normal, then all combinations are possible. the value for type may be "mean" only if the distribution is log-normal, and the only additional option is "estimate" for se.fit. For either log-normal or common log distributions, if the type is "response," then the options for se.fit and interval ae "none" or "prediction."

Value

A vector of predictions or a matrix of predictions and the other selected statistics. In contrast with other regression predict functions, a list of fitted values and standard errors is never returned.

Note

The predicted value for type = "mean" uses the adjusted maximum likelihood method for bias correction described by Cohn (1988) and Cohn and others (1992). that method will give sligthly different predicted values than the minimum variance unbiased estimate (MVUE) described by Bradu and Mundlak (1970).

References

Bradu, D., and Mundlak, Y. 1979, Estimation in lognormal linear models: Journal of the American Statistical Association, v. 65, no. 39, p. 198–211.
Cohn, T.A., 1988, Adjusted maximum likelihood estimation of the moments of lognormal populations form type I censored samples: U.S. Geological Survey Open-File Report 88-350, 34 p.
Cohn, T.A., Gilroy, E.J., and Baier, W.G., 1992, Estimating fluvial transport of trace constituents using a regression model with data subject to censoring: Proceeding of the Joint Statistical Meeting, Boston, August 9–13, 1992, p. 142–151.

See Also

censReg, predictMVUE

USGS-R/smwrQW documentation built on Oct. 11, 2022, 6:13 a.m.