doRetro: Run a retrospective evaluation.

In MichaelFolkes/forecastR_package: Forecast stuff

Description

Run a retrospective evaluation.

Usage

doRetro(
  model = NULL,
  data = NULL,
  predictors = NULL,
  covariates = NULL,
  retro.settings = NULL,
  fit.settings = NULL,
  fc.settings = NULL,
  tracing = FALSE,
  out.type = c("short", "full"),
  interval.n = 500,
  interval.quants = TRUE,
  pt.fc.in = NULL
)

Arguments

model	A character vector of length one. It is the model name and must match one of the model names in the estimation.functions list object.
data	A data frame. Equivalent to the "data" element in the list output from prepData.
predictors	A list. Optional input needed for return rate model. Equal to the element named predictors from output of prepData. Default is NULL.
covariates	A list. Optional input needed for covariate model. Equal to the element named covariates from output of prepData. Default is NULL.
retro.settings	A list
fit.settings	A list
fc.settings	A list
tracing	A Boolean (default is FALSE)
out.type	A character vector of length one. Can have value "short" (default) or "full". See value for more information.
interval.n	An integer. The sample size for generating the retro-based interval. Default is 500.
interval.quants	A Boolean (default is TRUE). If TRUE the output includes quantiles of the retro-based interval, rather than the full sample.
pt.fc

Value

Key output (with out.type="short") from this function includes two summary tables with performance measures from the dynamic retrospective: retro.pm.all: for each age class, use as many retrospective years as possible (i.e more for younger ages). retro.pm.bal: balanced output, using the same retrospective years for all age classes and the total. If out.type="full", then the output also includes the full model fit info for each retro year

MichaelFolkes/forecastR_package documentation built on April 4, 2021, 5:14 a.m.