pit_hist: Compute PIT Histogram

In robertsy/assimilr: Local Ensemble Kalman Particle Filters

Description

PIT= Probability integral transform, also called Rank Histogram

Usage

pit_hist(hx, y, fig = TRUE, ...)

Arguments

hx	ensemble prediction (dxK :one row for each y, one col for each ensemble member)
y	actual vector of observations
fig	should the histogram be plotted or the ranks be returned?

Value

the vector of ranks (length of y)

Examples

set.seed(1)
d <- 1000
yo <- rnorm(d)
k <- 40
pit <- pit_hist(matrix(rnorm(d*k, 0, 1  ), d, k), yo, main='calibrated ensemble')
pit <- pit_hist(matrix(rnorm(d*k, 0, 0.5), d, k), yo, main='underdispersed')
pit <- pit_hist(matrix(rnorm(d*k, 0, 1.5), d, k), yo, main='overdispersed')

robertsy/assimilr documentation built on May 27, 2019, 10:33 a.m.