StMoMo (version 0.4.1)

bootstrap.fitStMoMo: Bootstrap a fitted Stochastic Mortality Model

Description

Produce bootstrap parameters of a Stochastic Mortality Model to account for parameter uncertainty.

Usage

# S3 method for fitStMoMo
bootstrap(object, nBoot = 1, type = c("semiparametric",
  "residual"), deathType = c("observed", "fitted"), ...)

Arguments

object

an object of class "fitStMoMo" with the fitted parameters of a stochastic mortality model.

nBoot

number of bootstrap samples to produce.

type

type of bootstrapping approach to be applied. "semiparametric"(default) uses the assumed distribution of the deaths to generate bootstrap samples. "residual" resamples the deviance residuals of the model to generate bootstrap samples.

deathType

type of deaths to sample in the semiparametric bootstrap. "observed" (default) resamples the observed deaths. "fitted" resamples the fitted deaths. This parameter is only used if type is "semiparametric".

...

arguments to be passed to or from other methods.

Value

A list with class "bootStMoMo" with components:

bootParameters

a list of of length nBoot with the fitted parameters for each bootstrap replication.

model

the model fit that has been bootstrapped.

type

type of bootstrapping approach applied.

deathType

type of deaths sampled in case of semiparametric bootstrap.

Details

When type is "residual" the residual bootstrapping approach described in Renshaw and Haberman (2008) is applied, which is an adaptation of the approach of Koissi et al (2006). In the case of a "logit" link with Binomial responses the adaptation described in Debon et al, (2010, section 3) is used.

When type is "semiparametric" the semiparametric approach described in Brouhns et al.(2005) is used. In the case of a "logit" link with Binomial responses a suitable adaptation is applied. If deathType is "observed" then the observed deaths are used in the sampling as in Brouhns et al. (2005) while if deathType is "fitted" the fitted deaths are used in the sampling as in Renshaw and Haberman (2008).

References

Brouhns, N., Denuit M., & Van Keilegom, I. (2005). Bootstrapping the Poisson log-bilinear model for mortality forecasting. Scandinavian Actuarial Journal, 2005(3), 212-224.

Debon, A., Martinez-Ruiz, F., & Montes, F. (2010). A geostatistical approach for dynamic life tables: The effect of mortality on remaining lifetime and annuities. Insurance: Mathematics and Economics, 47(3), 327-336.

Renshaw, A. E., & Haberman, S. (2008). On simulation-based approaches to risk measurement in mortality with specific reference to Poisson Lee-Carter modelling. Insurance: Mathematics and Economics, 42(2), 797-816.

See Also

simulate.bootStMoMo, plot.bootStMoMo

Examples

# NOT RUN {
#Long computing times
# }
# NOT RUN {
LCfit <- fit(lc(), data = EWMaleData)

LCResBoot <- bootstrap(LCfit, nBoot = 500, type = "residual")
plot(LCResBoot)
LCSemiObsBoot <- bootstrap(LCfit, nBoot = 500, type = "semiparametric")
plot(LCSemiObsBoot)
LCSemiFitBoot <- bootstrap(LCfit, nBoot = 500, type = "semiparametric",  
                           deathType = "fitted")
plot(LCSemiFitBoot)
# }
# NOT RUN {
# }