Read a population total off an estimate

A death rate is a ratio of two visibility-adjusted sums. Those sums are estimands in their own right, and this reads one of them out with its uncertainty.

Usage

estimated_total(res, of = c("events", "exposure"), estimator = c("ind", "agg"))

Arguments

res

a result from network_survival_estimator()

of

"events" for the numerator — deaths, for a mortality estimate — or "exposure" for the denominator, the estimated person-time

estimator

"ind" (individual visibility) or "agg" (aggregate)

Value

a tibble with one row per cell: the estimate, and its interval and standard error where the estimate was bootstrapped

Why this is a separate step

network_survival_estimator() computes visibility-adjusted weighted sums and then divides one by the other. The division is the last thing it does, and nothing before it assumes a ratio is what you want — the visibility layer in particular has no idea what is being summed. So a total needs no new estimation, only a different final step.

What makes it a population total

The sum is only an estimate of a population total if the survey weights are population weights. With relative or normalised weights it is a total on whatever scale those weights carry, and the number will be wrong by a constant factor — silently, since nothing in the data says which kind of weight was used. The rate is unaffected either way, because the factor cancels; that is exactly why this distinction can go unnoticed until somebody asks for a total.

See also

network_survival_estimator()

Examples

if (FALSE) { # \dontrun{
estimated_total(res, "events")     # estimated deaths in the population
estimated_total(res, "exposure")   # estimated person-time
} # }