maple: maple: model averaged projections of life expectancy maple is...
In vkontis/maple: Model averaged projections of life expectancy

Description Usage Arguments Value Note Examples

maple: model averaged projections of life expectancy maple is an R package for fitting statistical models on demographic data to produce projections of death rates and life expectancy.

Produce model averaged projections of death rates and life expectancy.

1
2
3

maple(deaths, population, forecast.horizon, holdout, models = maple_models(),
  num.draws = 1000, ax = NULL,
  num.threads = inla.getOption("num.threads"), verbose = TRUE)

`deaths`	A matrix of death counts, with 18 rows, one for each 5-year age group 0-4, ..., 80-84, 85+ and one column for each year of available data. The column names of the matrix must be the years of data.
`population`	A matrix of mid-year population numbers, in the same row/column format as deaths.
`forecast.horizon`	The number of years to produce projections for.
`holdout`	The number of years of data to hold out to calculate model weights.
`models`	The individual models to be run and averaged; see ?maple_models for more details.
`num.draws`	The number of posterior samples from individual model fits to use for producing the BMA estimates.
`ax`	The number of years lived on average by those who die in their current age group. See ?maple_plt for more details.
`num.threads`	The number of threads to use when running the models. This is passed to the INLA methods. If not specified, then all available threads are used.
`verbose`	If TRUE (the default), print some information on progress fitting models, etc.

A list with the following entries

model.weights: The weights used to combine models into the model average.
sample.summaries: A data frame holding statistical summary information for age-specific death rates, life expectancy and probability of dying, calculated from the posterior draws.
samples: A list of life table draws, calculated using posterior samples of death rates.
individual.model.sample.summaries: A data frame containing predictions under individual models.

The maximum possible number of BMA samples is taken, depending on the model weights. For example if num.draws == 1000 and there are 5 models with weights 0.25, 0.2, 0.2, 0.2, and 0.15, the code will try to use all draws from the first model (with largest weight) and a number of draws from the remaining models inversely proportional to their weights (800, 800, 800, 600).

data(maple.deaths)
data(maple.population)
data(maple.ax)
models <- maple_models()[c(1, 20)]
bma <- maple(deaths = maple.deaths, population = maple.population,
             forecast.horizon = 20, holdout = 13, models = models,
             ax = maple.ax)