Leave-One-Out Influence on Population Size

Computes the change in the total estimated population size \(\hat\xi = \sum_i N_i^{\hat\alpha_i}\) when each observation (or country) is dropped. This is a convenience wrapper around loo.

Usage

dfpopsize(model, ...)

# S3 method for class 'uncounted'
dfpopsize(model, by = c("obs", "country"), ...)

Arguments

model

A fitted uncounted object.

...

Additional arguments passed to loo.

by

Character: "obs" (default) or "country".

Value

A named numeric vector. Each element is the change in \(\hat\xi\) when that unit is dropped (\(\hat\xi_{(-i)} - \hat\xi\)).

See also

loo, dfbeta.uncounted

Examples

data(irregular_migration)
d <- irregular_migration[irregular_migration$year == "2019" & irregular_migration$sex == "Male", ]
fit <- estimate_hidden_pop(d, ~ m, ~ n, ~ N, method = "poisson",
                           gamma = 0.001, countries = ~ country_code)
dp <- dfpopsize(fit)
head(dp)
#>          1          2          3          4          5          6 
#> 4006.95264 -157.03916 3460.55759  187.08577  542.66030  -30.66062