collapseIterations: Summarize iterations.

In StatisticsNZ/dembase: Analysing Cross-Classified Data about Populations

Description

Collapse a dimension with dimtype "iteration", typically replacing it with a much shorter dimension with dimtype "quantile".

Usage

collapseIterations(object, FUN = quantile, ...)

## S4 method for signature 'DemographicArray'
collapseIterations(object, FUN = quantile, ...)

Arguments

object	Object of class DemographicArray.
FUN	A function. Defaults to quantile, though others can be used.
...	Further arguments, passed to FUN.

Details

Many estimation and forecasting methods produce sets of simulation runs or iterations. When graphing or reporting results, it is typically necessary to reduce the full set of iterations to a set of summary values. Function collapseIterations carries out this reduction. By default, collapseIterations calculates quantiles. However, users can supply alternatives, including their own customized functions.

If FUN returns a vector, then the dimension with dimtype "iteration" is replaced by one with dimtype "quantile" or "state". If FUN returns a single value, then the "iteration" dimension is dropped.

By default, function quantile uses quantiles 0, 0.25, 0.5, 0.75, 1. Quantiles more useful for summarizing iterations, such as 0.025, 0.5, 0.975, can be obtained by using a prob argument: see below for an example.

See Also

collapseCategories, collapseIntervals, collapseOrigDest, collapseDimension, dimtypes, dimscales

Examples

x <- Counts(array(rpois(n = 20, lambda = c(8, 12)),
                  dim = c(2, 10),
                  dimnames = list(sex = c("Female", "Male"),
                      iteration = 1:10)))
x
collapseIterations(x)

## pass different values for 'prob' to 'quantile'
collapseIterations(x, prob = c(0.025, 0.5, 0.975))

## summarize with a single value
collapseIterations(x, FUN = mean)

## a user-defined function
sumvals <- function(x) c(mean = mean(x), sd = sd(x))
collapseIterations(x, FUN = sumvals)

StatisticsNZ/dembase documentation built on Dec. 25, 2021, 4:49 p.m.