BAS (version 1.7.3)

predict.bas: Prediction Method for an object of class BAS

Description

Predictions under model averaging or other estimators from a BMA object of class inheriting from 'bas'.

Usage

# S3 method for bas
predict(
  object,
  newdata,
  se.fit = FALSE,
  type = "link",
  top = NULL,
  estimator = "BMA",
  na.action = na.pass,
  ...
)

Value

a list of

fit

fitted values based on the selected estimator

Ybma

predictions using BMA, the same as fit for non-BMA methods for compatibility; will be deprecated

Ypred

matrix of predictions under each model for BMA

se.fit

se of fitted values; in the case of BMA this will be a matrix

se.pred

se for predicted values; in the case of BMA this will be a matrix

se.bma.fit

vector of posterior sd under BMA for posterior mean of the regression function. This will be NULL if estimator is not 'BMA'

se.bma.pred

vector of posterior sd under BMA for posterior predictive values. this will be NULL if estimator is not 'BMA'

best

index of top models included

bestmodels

subset of bestmodels used for fitting or prediction

best.vars

names of variables in the top model; NULL if estimator='BMA'

df

scalar or vector of degrees of freedom for models

estimator

estimator upon which 'fit' is based.

Arguments

object

An object of class BAS, created by bas

newdata

dataframe for predictions. If missing, then use the dataframe used for fitting for obtaining fitted and predicted values.

se.fit

indicator for whether to compute se of fitted and predicted values

type

Type of predictions required. "link" which is on the scale of the linear predictor is the only option currently for linear models, which for the normal model is equivalent to type='response'.

top

a scalar integer M. If supplied, subset the top M models, based on posterior probabilities for model predictions and BMA.

estimator

estimator used for predictions. Currently supported options include:
'HPM' the highest probability model
'BMA' Bayesian model averaging, using optionally only the 'top' models
'MPM' the median probability model of Barbieri and Berger.
'BPM' the model that is closest to BMA predictions under squared error loss. BMA may be computed using only the 'top' models if supplied

na.action

function determining what should be done with missing values in newdata. The default is to predict NA.

...

optional extra arguments

Author

Merlise Clyde

Details

Use BMA and/or model selection to form predictions using the top highest probability models.

See Also

bas, fitted.bas, confint.pred.bas, variable.names.pred.bas

Other predict methods: fitted.bas(), predict.basglm(), variable.names.pred.bas()

Other bas methods: BAS, bas.lm(), coef.bas(), confint.coef.bas(), confint.pred.bas(), diagnostics(), fitted.bas(), force.heredity.bas(), image.bas(), plot.confint.bas(), predict.basglm(), summary.bas(), update.bas(), variable.names.pred.bas()

Examples

data("Hald")
hald.gprior =  bas.lm(Y ~ ., data=Hald, alpha=13, prior="g-prior")

predict(hald.gprior, newdata=Hald, estimator="BPM", se.fit=TRUE)
# same as fitted
fitted(hald.gprior,estimator="BPM")
# default is BMA and estimation of mean vector
hald.bma = predict(hald.gprior, top=5, se.fit=TRUE)
confint(hald.bma)

hald.bpm = predict(hald.gprior, newdata=Hald[1,],
                    se.fit=TRUE,
                    estimator="BPM")
confint(hald.bpm)
# extract variables
variable.names(hald.bpm)

hald.hpm = predict(hald.gprior, newdata=Hald[1,],
                    se.fit=TRUE,
                    estimator="HPM")
confint(hald.hpm)
variable.names(hald.hpm)

hald.mpm = predict(hald.gprior, newdata=Hald[1,],
                    se.fit=TRUE,
                    estimator="MPM")
confint(hald.mpm)
variable.names(hald.mpm)