model.Oeppen: The Oeppen Mortality Model (Oeppen - CoDa)

In mpascariu/MortalityForecast: Standard Tools to Compare and Evaluate Various Mortality Forecasting Methods

Description

Fit the Oeppen model for forecasting the life table distribution of deaths. This is a Lee-Carter type model adapted to a compositional-data framework (CoDa). A key difference between the \insertCitelee1992;textualMortalityForecast method and the Oeppen model is that the former fits and forecasts the death rates (mx) while the latter is based on the life table death distribution (dx). \insertCite@See @oeppen2008 and @bergeron2017;textualMortalityForecast for a detail description and mathematical formulation.

Usage

model.Oeppen(data, x = NULL, y = NULL, verbose = TRUE, ...)

Arguments

data	A data.frame or a matrix containing mortality data with ages x as row and time y as column.
x	Numerical vector indicating the ages in input data. Optional. Default: NULL.
y	Numerical vector indicating the years in input data. Optional. Default: NULL.
verbose	A logical value. Set verbose = FALSE to silent the process that take place inside the function and avoid progress messages.
...	Arguments to be passed to or from other methods.

Value

The output is a list with the components:

input	List with arguments provided in input. Saved for convenience;
info	Short details about the model;
call	An unevaluated function call, that is, an unevaluated expression which consists of the named function applied to the given arguments;
coefficients	Estimated coefficients;
fitted.values	Fitted values of the estimated model;
observed.values	The observed values used in fitting arranged in the same format as the fitted.values;
residuals	Deviance residuals;
x	Vector of ages used in the fitting;
y	Vector of years used in the fitting.

Author(s)

Marius D. Pascariu and Marie-Pier Bergeron-Boucher

References

\insertAllCited

See Also

predict.Oeppen plot.Oeppen

Examples

# Example 1 ----------------------
# Data
x  <- 0:100
y  <- 1980:2016
dx <- HMD_male$dx$GBRTENW[paste(x), paste(y)]

# If data contains zero's we have to replace them with very small 
# values in order to avoid errors in fitting. replace.zeros() will do it.
dx <- replace.zeros(dx)

# Fit model
M <- model.Oeppen(data = dx, x = x, y = y)
M

summary(M)
coef(M)

# Plot observed and fitted values
plot(M, plotType = "observed")
plot(M, plotType = "fitted")

# Plot residuals
R <- residuals(M)
plot(R, plotType = "scatter")
plot(R, plotType = "colourmap")
plot(R, plotType = "signplot")

# Perform forecasts
P  <- predict(M, h = 16)
P

plot(P, plotType = "mean")
plot(P, plotType = "lower")
plot(P, plotType = "upper")

#' # Example 2 ----------------------
# One can specify manually the ARIMA order, a drift to be included or not, 
# and the jump choice of the first forecast year.
P2 <- predict(M, h = 20, 
              order = c(0,1,1), 
              include.drift = FALSE, 
              jumpchoice = "fit")

## Not run: 
# Example 3 ----------------------
# Compute life tables using forecast values using the MortalityLaws R package
library(MortalityLaws)
dx <- P$predicted.values
lt <- LifeTable(x = P$x, dx = dx)

## End(Not run)

mpascariu/MortalityForecast documentation built on Sept. 28, 2020, 2:40 p.m.