Functional mean model

FMEAN() returns an iid functional model applied to the formula's response variable as a function of age. Standard deviations that cannot be estimated, such as at ages with fewer than two finite residuals, are interpolated from neighbouring ages. Simulations from generate() with bootstrap = TRUE resample whole years of residuals, so they keep the correlation between ages. Otherwise, ages are simulated independently from normal distributions, which understates the uncertainty of quantities computed across ages, such as life expectancy.

Usage

FMEAN(formula, ...)

Arguments

formula

Model specification.

...

Not used. An error is given if any arguments are supplied here, so that misspelled arguments are not silently ignored.

Value

A model specification.

Author

Rob J Hyndman

Examples

fmean <- norway_mortality |>
  dplyr::filter(Sex == "Female") |>
  model(mean = FMEAN(Mortality))
report(fmean)
#> Series: Mortality 
#> Model: FMEAN 
#> 
#> # A tibble: 111 × 3
#>      Age     mean   sigma
#>    <int>    <dbl>   <dbl>
#>  1     0 0.0231   0.0212 
#>  2     1 0.00616  0.00763
#>  3     2 0.00262  0.00324
#>  4     3 0.00179  0.00215
#>  5     4 0.00144  0.00173
#>  6     5 0.00124  0.00152
#>  7     6 0.00109  0.00134
#>  8     7 0.000972 0.00118
#>  9     8 0.000907 0.00111
#> 10     9 0.000879 0.00113
#> # ℹ 101 more rows
autoplot(fmean) + ggplot2::scale_y_log10()