DoubleGap {MortalityGaps}

Fit the DoubleGap Life Expectancy Forecasting Model

Description

Fit a DoubleGap model for forecasting life expectancy. The method combines separate forecasts to obtain joint male and female life expectancies that are coherent with a best-practice trend. See the entire description and mathematical formulation of the model in Pascariu et. al (2017).

Usage

DoubleGap(DF, DM, age, country, years, arima.order = NULL, drift = NULL,
  tau = NULL, A = NULL)

Arguments

DF

data.frame containing life expectancy records for females. The table must contain the following 4 columns: country, year, age, ex.

DM

data.frame containing life expectancy records for males. The table must have the same format and dimensions as DF.

age

Indicate the age for which the model to be fitted. Assuming DF and DM contain recods for different ages, this argument it is used to subset the data. If you want to fit the model for age 0, add age = 0. Type: scalar.

country

Indicate for which contry you want to fit the model. The country name or code must exist in DF and DM. Type: character.

years

Period of time to be used. Type: numeric vector.

arima.order

A specification of the the ARIMA model to be used in fitting the best-practice gap. The ARIMA order is country specific. The three integer components (p, d, q) are the AR order, the degree of differencing, and the MA order. Format: numerical vector of length 3. If arima.order = NULL the function conducts a search over possible models according to AIC. See auto.arima for details.

drift

Indicate whether the ARIMA model should include a linear drift term or not. Type: logical value. If drift = NULL, it will be estimate automatically.

tau

The level of female life expectancy at which the sex-gap is expected to stop widening and to start narrowing. If NULL then the model will run an algorithm to find it.

A

The level of female life expectancy where we assume no further change in the sex-gap. If NULL the model will estimate it.

Value

The output is of "DoubleGap" class with the components:

input

List with arguments provided in input. Saved for convenience.

call

Short information about the model.

coefficients

Estimated coefficients.

fitted.values

Fitted values of the selected model.

observed.values

Country specific observed values.

model.parts

Object containing detailed results of the fitted model.

residuals

Deviance residuals.

Author(s)

Marius D. Pascariu

References

Pascariu M.D., Canudas-Romo V. and Vaupel W.J. 2017. The double-gap life expectancy forecasting model. Insurance: Mathematics and Economics Volume 78, January 2018, Pages 339-350.

See Also

predict.DoubleGap

Examples

# Input data ------------------------------------
# Collection of life expectancies for female populations
exF <- MortalityGaps.data$exF
# Life expectancy for male populations
exM <- MortalityGaps.data$exM

# Example 1 ----------------------------------------------
# Fit DG model at age 0 for Australia using data from 1950 to 2014
M0 <- DoubleGap(DF = exF,
                DM = exM,
                age = 0,
                country = "AUS",
                years = 1950:2014)
M0
summary(M0)
ls(M0)

# Forecast life expectancy in Australia until 2030
P0 <- predict(M0, h = 16)
P0
# Plot the results
plot(P0)

## Not run: 
# Example 2 ----------------------------------------------
# Fit DG model at age 0 for Sweden. Provide details about models.
# Reproduce published results in the article.
M1 <- DoubleGap(DF = exF, 
                DM = exM, 
                age = 0, 
                country = "SWE", 
                years = 1950:2014, 
                arima.order = c(2, 1, 1), 
                drift = TRUE, 
                tau = 75, 
                A = 86)
summary(M1)
# Predict model 
P1 <- predict(M1, h = 36)
plot(P1)

# Example 3 ----------------------------------------------
# Fit DG model for USA at age 65.
M2 <- DoubleGap(DF = exF, 
                DM = exM, 
                age = 65, 
                country = "USA", 
                years = 1950:2014, 
                arima.order = c(0, 1, 0), 
                drift = FALSE, 
                tau = 15, 
                A = 24)
summary(M2)
# Predict model 
P2 <- predict(M2, h = 36)
plot(P2)

## End(Not run)