actuar (version 3.3-4)

rcomphierarc.summaries: Summary Statistics of a Portfolio

Description

Methods for class "portfolio" objects.

aggregate splits portfolio data into subsets and computes summary statistics for each.

frequency computes the frequency of claims for subsets of portfolio data.

severity extracts the individual claim amounts.

weights extracts the matrix of weights.

Usage

# S3 method for portfolio
aggregate(x, by = names(x$nodes), FUN = sum,
        classification = TRUE, prefix = NULL, ...)
# S3 method for portfolio
frequency(x, by = names(x$nodes),
        classification = TRUE, prefix = NULL, ...)
# S3 method for portfolio
severity(x, by = head(names(x$node), -1), splitcol = NULL,
        classification = TRUE, prefix = NULL, ...)
# S3 method for portfolio
weights(object, classification = TRUE, prefix = NULL, ...)

Value

A matrix or vector depending on the groupings specified in by.

For the aggregate and frequency methods: if at least one level other than the last one is used for grouping, the result is a matrix obtained by binding the appropriate node identifiers extracted from x$classification if classification = TRUE, and the summaries per grouping. If the last level is used for grouping, the column names of x$data are retained; if the last level is not used for grouping, the column name is replaced by the deparsed name of

FUN. If only the last level is used (column summaries), a named vector is returned.

For the severity method: a list of two elements:

main

NULL or a matrix of claim amounts for the columns not specified in splitcol, with the appropriate node identifiers extracted from x$classification if classification = TRUE;

split

same as above, but for the columns specified in splitcol.

For the weights method: the weight matrix of the portfolio with node identifiers if classification = TRUE.

Arguments

x, object

an object of class "portfolio", typically created with simul.

by

character vector of grouping elements using the level names of the portfolio in x. The names can be abbreviated.

FUN

the function to be applied to data subsets.

classification

boolean; if TRUE, the node identifier columns are included in the output.

prefix

characters to prefix column names with; if NULL, sensible defaults are used when appropriate.

splitcol

columns of the data matrix to extract separately; usual matrix indexing methods are supported.

...

optional arguments to FUN, or passed to or from other methods.

Author

Vincent Goulet , Louis-Philippe Pouliot.

Details

By default, aggregate.portfolio computes the aggregate claim amounts for the grouping specified in by. Any other statistic based on the individual claim amounts can be used through argument FUN.

frequency.portfolio is equivalent to using aggregate.portfolio with argument FUN equal to if (identical(x, NA)) NA else length(x).

severity.portfolio extracts individual claim amounts of a portfolio by groupings using the default method of severity. Argument splitcol allows to get the individual claim amounts of specific columns separately.

weights.portfolio extracts the weight matrix of a portfolio.

See Also

rcomphierarc

Examples

nodes <- list(sector = 3, unit = c(3, 4),
              employer = c(3, 4, 3, 4, 2, 3, 4), year = 5)
model.freq <- expression(sector = rexp(1),
                         unit = rexp(sector),
                         employer = rgamma(unit, 1),
                         year = rpois(employer))
model.sev <- expression(sector = rnorm(6, 0.1),
                        unit = rnorm(sector, 1),
                        employer = rnorm(unit, 1),
                        year = rlnorm(employer, 1))
pf <- rcomphierarc(nodes, model.freq, model.sev)

aggregate(pf)            # aggregate claim amount by employer and year
aggregate(pf, classification = FALSE) # same, without node identifiers
aggregate(pf, by = "sector")	      # by sector
aggregate(pf, by = "y")		      # by year
aggregate(pf, by = c("s", "u"), mean) # average claim amount

frequency(pf)			      # number of claims
frequency(pf, prefix = "freq.")       # more explicit column names

severity(pf)			      # claim amounts by row
severity(pf, by = "year")	      # by column
severity(pf, by = c("s", "u"))        # by unit
severity(pf, splitcol = "year.5")     # last year separate
severity(pf, splitcol = 5)            # same
severity(pf, splitcol = c(FALSE, FALSE, FALSE, FALSE, TRUE)) # same

weights(pf)

## For portfolios with weights, the following computes loss ratios.
if (FALSE) aggregate(pf, classif = FALSE) / weights(pf, classif = FALSE)