forecast method for spdur class objects.
# S3 method for spdur forecast(object, ..., pred.data = NULL, stat = "conditional hazard", n.ahead = 6)
| object | A |
|---|---|
| … | Optional arguments, not used. |
| pred.data | Data on which to base forecasts, i.e. slice of last time unit's observations for all cross-sectional units. |
| stat | Which statistic to forecast, see |
| n.ahead | How many time periods to predict ahead. Default is 6. |
This function will create out-of-sample predictions of ``stat''
using model estimates and the prediction data provided. It is assumed that
prediction data consist of a slice of the last time period observed for
the data used to estimate the model in object. For each row,
forecast.spdur will estimate the model predictions for that time point
and then extrapolate the resulting probability to n.ahead time
periods using appropriate probability theory.
For situations in which the covariate values are known for future time
periods, e.g. in a test sample use predict.spdur instead.
library(forecast) data(coups) data(model.coups) coups.dur <- add_duration(coups, "succ.coup", "gwcode", "year", freq="year") pred.data <- coups.dur[coups.dur$year==max(coups.dur$year), ] pred.data <- pred.data[complete.cases(pred.data), ] fcast <- forecast(model.coups, pred.data=pred.data)