Skip to contents

Peaks-over-threshold point process model

Source: R/model.R
model_point_process.Rd

A point process over exceedances of a threshold, combining an arrival rate with a magnitude distribution.

Usage

model_point_process(
  data,
  threshold = NULL,
  total_years = NULL,
  use_defaults = TRUE,
  seasonal = FALSE,
  time_block = c("water_year", "calendar_year", "custom_year", "quarter", "month"),
  start_month = 10,
  parameters = NULL,
  parameter_values = NULL,
  use_default_flat_priors = NULL
)

Arguments

data

a numeric vector of observations, or an analysis_data() frame.

threshold

optional exceedance threshold; derived from the data when omitted. See threshold_diagnostics() for choosing one.

total_years

optional record length in years, used for the arrival rate.

use_defaults

logical; let the model derive its threshold and record length from the data. Applied before an explicit threshold or total_years, so an explicit value wins.

seasonal

logical; fit two seasonal magnitude distributions with fitted change points.

time_block

the block a seasonal model reduces the record over: "water_year" (the default), "calendar_year", "custom_year", "quarter" or "month".

start_month

the month a water year or custom year begins. Default 10 (October).

parameters

optional model_parameter() object or list of them, setting bounds, fixed flags, priors, or starting values.

parameter_values

optional numeric vector of all parameter values, applied last.

use_default_flat_priors

logical; set FALSE alongside a custom prior.

Value

An object of class corehydro_model.

Details

With seasonal = TRUE the magnitude distribution becomes TWO generalized extreme value marginals with two fitted change points, and each observation is assigned to a season by its day of the year – so the data must carry DATES rather than bare indices:

data <- analysis_data(exact = list(date = dates, value = peaks))
model <- model_point_process(data, seasonal = TRUE)

Observations supplied with integer indices are treated as January 1 of that year, which puts every one of them in the same season.

See also

point_process_analysis(), threshold_diagnostics().

Examples

x <- c(1200, 1500, 890, 2210, 1870, 1420, 980, 2850, 1740, 1160, 1920, 1380, 2560, 1050)
model_point_process(x, threshold = 1000, total_years = 14)
#> <corehydro_model> point_process
#>   data: 14 exact observations