BAPC: Function to project age-specific mortality or disease rates...

In BAPC: Projection of cancer incidence and mortality data using Bayesian APC models fitted with INLA

Function to project age-specific mortality or disease rates using INLA

Description

A Bayesian age-period-cohort model fitted using a Poisson model within INLA is used to project mortality or disease rates. Age, period and/or cohort effects are either modelled using a random walk of second order (RW2), or fixed effect (drift).

Usage

  BAPC(APCList, predict = list(npredict = 0, retro = TRUE),
  model = list(age = list(model = "rw2", prior="loggamma", param = c(1, 0.00005), initial = 4, scale.model=FALSE),
    period = list(include = TRUE, model = "rw2", prior="loggamma", param = c(1, 0.00005), initial = 4, scale.model=FALSE),
    cohort = list(include = TRUE, model = "rw2", prior="loggamma", param = c(1, 0.00005), initial = 4, scale.model=FALSE),
    overdis = list(include = TRUE, model = "iid", prior="loggamma", param = c(1, 0.005), initial = 4)),
    secondDiff = FALSE, stdweight = NULL, verbose = FALSE)

Arguments

APCList	a APCList object
predict	a list specifying how many periods are to be projected and whether existing counts shoud be removed and projected. The first argument npredict sets the number of periods to predict. The second argument retro is a boolean indicating whether the past npredict periods should be removed or predicted or whether future rates are to be predicted. If retro is FALSE data cells for which the corresponing observation count is set to NA will be projected.
model	a list composed of four arguments: age, period, cohort, overdis. For each argument a separate list is to be specified defining whether the component should be included and if so, which model should be used. Possible arguments are: include A Boolean indictating whether the component should be included in the model. Age is per default always included. model A character indicating the model to be used. Possible choices are rw2 and drift. prior Prior distribution for the (log) precision parameter, such as "loggamma" or "pc.prec" for the penalised-complexity prior. param Parameters for the prior distribution for the precision parameter. initial Initial value for the precision parameter (on log-scale). scale.model Logical. If TRUE then scale the RW1 and RW2 models so that their (generlized) variance is 1, see Sorbye and Rue (2014).
secondDiff	Boolean (default:FALSE) indicating whether summary estimates for the second differences of age, period and cohort effects should be computed.
stdweight	Numeric vector with length equal to the number of age groups used to derive age-standardized projections. If the weights do not sum to one, they will be normalised internally. If no weights are provided, there will no age-standardized projections be computed.
verbose	Boolean (default:FALSE) indicating whether the program should run in a verbose model.

Value

An APCList object.

Author(s)

Andrea Riebler and Leonhard Held

References

Havard Rue, Sara Martino, and Nicholas Chopin (2009). Approximate Bayesian Inference for Latent Gaussian Models Using Integrated Nested Laplace Approximations. Journal of the Royal Statistical Society B, 71, 319-392. (www.r-inla.org)

Sigrunn Holbek Sorbye and Havard Rue (2014). "Scaling intrinsic Gaussian Markov random field priors in spatial modelling." Spatial Statistics 8: 39-51.

See Also

inla

Examples

## Not run: 
if(requireNamespace("INLA", quietly = TRUE)) {
   require(INLA)
   data(FemLCSweden)
   data(FemPYSweden)
   data(whostandard)

   lc_sweden <- APCList(FemLCSweden, FemPYSweden, gf=5)

   result <- BAPC(lc_sweden, predict=list(npredict=10, retro=TRUE),
      secondDiff=FALSE, stdweight=whostandard[6:17,2], verbose=FALSE)
}

## End(Not run)

BAPC documentation built on March 23, 2022, 3 p.m.