simITS (version 0.1.1)

extrapolate_model: Extrapolate pre-policy data to post-policy era

Description

This function takes a fitted model and uses it to make the post-policy predictions by simulating data.

Usage

extrapolate_model(
  M0,
  outcomename,
  dat,
  t0,
  R = 400,
  summarize = FALSE,
  smooth = FALSE,
  smoother = smooth_series,
  full_output = FALSE,
  fix_parameters = FALSE,
  ...
)

Arguments

M0

The fit model

outcomename

Outcome of interest (name of column)

dat

Dataframe with data being analyzed.

t0

Last pre-policy timepoint

R

Number of replications

summarize

Boolean, TRUE means collapse all simulated trajectories into single aggregate. FALSE means return all paths.

smooth

Boolean. TRUE means fit a smoother to the trajectories and look at distribution of smoothed trajectories. FALSE means look at raw data treajectories.

smoother

Function to do smoothing, if smoothing set to TRUE. Default is smooth_series()

full_output

TRUE means smoother returns residuals as well as smoothed series.

fix_parameters

Keep the parameters in the model M0 as fixed; do not add parameter uncertainty.

...

Extra arguments to be passed to smoother (e.g, bandwidth).

Value

Dataframe with columns corresponding to the simulations. If summarize=TRUE, one row per month in original data. If FALSE, all the details of all the runs are returned.

See Also

process_outcome_model for wrapper function for this method that is easier to use.

Examples

# NOT RUN {
data("mecklenberg" )
mecklenberg = add_lagged_covariates( mecklenberg, "pbail"  )
mecklenberg.pre = dplyr::filter( mecklenberg, month <= 0 )
M0 = fit_model_default( mecklenberg.pre, "pbail" )
res = extrapolate_model( M0, "pbail", mecklenberg, 0, 1,
                         smooth=TRUE)
tail( res )
# }