predict.plm: Model Prediction for plm Objects

Description

Predicted values of response based on plm models.

Usage

# S3 method for plm
predict(
  object,
  newdata = NULL,
  na.fill = !inherits(newdata, "pdata.frame"),
  ...
)

Value

A numeric (or a pseries if newdata is a pdata.frame) carrying the predicted values with length equal to the number of rows as the data supplied in newdata and with names the row names of newdata or, if newdata = NULL, the fitted values the original model given in object.

Arguments

object: An object of class "plm",
newdata: An optional pdata.frame in which to look for variables to be used for prediction. If NULL, the fitted values are returned. For fixed effects models, supplying a pdata.frame is recommended.
na.fill: A logical, only relevant if object is a pdata.frame, indicating whether for any supplied out-of-sample indexes (individual, time, combination of both), the missing fixed effect estimate is filled with the weighted mean of the model's present fixed effect estimates or not.
...: further arguments.

Details

predictcalculates predicted values by evaluating the regression function of a plm model for newdata or, if newdata = NULL, it returns the fitted values the plm model.

The fixed effects (within) model is somewhat special in prediction as it has fixed effects estimated per individual, time period (one-way) or both (two-ways model) which should to be respected when predicting values relating to these fixed effects in the model: To do so, it is recommended to supply a pdata.frame (and not a plain data.frame) in newdata as it describes the relationship between the data supplied to the individual. and/or time periods. In case the newdata´'s pdata.frame has out-of-sample data (data contains individuals and/or time periods not contained in the original model), it is not clear how values are to be predicted and the result will contain NA values for these out-of-sample data. Argument na.fill can be set to TRUE to apply the original model's weighted mean of fixed effects for the out-of-sample data to derive a prediction.

If a plain data.frame is given in newdata for a fixed effects model, the weighted mean is used for all fixed effects as newdata for prediction as a plain data.frame cannot describe any relation to individuals/time periods (na.fill is automatically set to TRUE and the function warns).

Examples

Run this code

library(plm)
data("Grunfeld", package = "plm")

# fit a fixed effect model
fit.fe <- plm(inv ~ value + capital, data = Grunfeld, model = "within")

# generate 55 new observations of three firms used for prediction:
#  * firm 1 with years 1935:1964 (has out-of-sample years 1955:1964), 
#  * firm 2 with years 1935:1949 (all in sample),
#  * firm 11 with years 1935:1944 (firm 11 is out-of-sample)
set.seed(42L)

new.value2   <- runif(55, min = min(Grunfeld$value),   max = max(Grunfeld$value))
new.capital2 <- runif(55, min = min(Grunfeld$capital), max = max(Grunfeld$capital))

newdata <- data.frame(firm = c(rep(1, 30), rep(2, 15), rep(11, 10)),
                      year = c(1935:(1935+29), 1935:(1935+14), 1935:(1935+9)),
                      value = new.value2, capital = new.capital2)
# make pdata.frame
newdata.p <- pdata.frame(newdata, index = c("firm", "year"))

## predict from fixed effect model with new data as pdata.frame
predict(fit.fe, newdata = newdata.p)

## set na.fill = TRUE to have the weighted mean used to for fixed effects -> no NA values
predict(fit.fe, newdata = newdata.p, na.fill = TRUE)

## predict with plain data.frame from fixed effect model: uses mean fixed effects 
## for prediction and, thus, yields different result with a warning
predict(fit.fe, newdata = newdata)