drnorm: Distributional random number generation (equivalent to rnorm)

In rothenhaeusler/calinf: Calibrated inference: inference that accounts for both sampling and distributional uncertainty

Distributional random number generation (equivalent to rnorm)

Description

Distributional random number generation (equivalent to rnorm)

Usage

drnorm(d_seed, mean = 0, sd = 1)

Arguments

d_seed	A distributional seed as returned by distributional_seed()
mean	vector of means of the unperturbed Gaussian distribution
sd	vector of standard deviations of the unperturbed Gaussian distribution

Examples

d_seed <- distributional_seed(1000,10)
qqnorm(drnorm(d_seed))
# Compare to standard normal
qqnorm(rnorm(1000))

# Test whether it works for multi-dimensional data generation
n <- 1000
delta <- 1.3
draw_perturbed_data <- function(n,delta){
  #Set distributional seed
  d_seed <- distributional_seed(n,delta)
  #Draw from perturbed model
  x <- drnorm(d_seed)
  y <- drnorm(d_seed)
  c(mean(x),mean(x^2),mean(x^3),mean(y),mean(y^2),mean(x*y))
}
mat <- sapply(rep(n,10000),draw_perturbed_data,delta=delta)
vec <- apply(mat,1,var)
# target delta^2
delta^2
# achieved delta^2 on simulated data
n*vec[1]/var(rnorm(10000))
n*vec[2]/var(rnorm(10000)^2)
n*vec[3]/var(rnorm(10000)^3)
n*vec[4]/var(rnorm(10000))
n*vec[5]/var(rnorm(10000)^2)
n*vec[6]/var(rnorm(10000)*rnorm(10000))

rothenhaeusler/calinf documentation built on May 2, 2022, 12:44 p.m.