simulAr1Ar1: AR1xAR1 simulation

In timflutre/rutilstimflutre: Timothee Flutre's personal R code

AR1xAR1 simulation

Description

Simulate random samples from a separable AR1xAR1 process as in Martin (1996).

Usage

simulAr1Ar1(
  n = 1,
  R = 2,
  C = 2,
  rho.r = 0,
  rho.c = 0,
  sigma.X.2 = 1,
  sigma.e.2 = 1
)

Arguments

n	number of samples
R	number of rows
C	number of columns
rho.r	correlation between rows
rho.c	correlation between columns
sigma.X.2	variance of X (see Martin, 1996, page 400)
sigma.e.2	variance of epsilons (see Martin, 1996, page 400)

Value

array which first dimension is R, second is C and third is n

Author(s)

Timothee Flutre

Examples

## Not run: ## strong correlation only between rows
set.seed(1234)
samples <- simulAr1Ar1(n=100, R=40, C=45, rho.r=0.8, rho.c=0)
dim(samples)
stats <- list()
stats$cor.btw.rows <- c(apply(samples, 3, function(mat){
  apply(mat, 2, function(row.i){ # per column
    acf(row.i, lag.max=1, type="correlation", plot=FALSE)$acf[2]
})}))
stats$cor.btw.cols <- c(apply(samples, 3, function(mat){
  apply(mat, 1, function(col.j){ # per row
    acf(col.j, lag.max=1, type="correlation", plot=FALSE)$acf[2]
})}))
sapply(stats, summary)

## strong correlation only between columns
set.seed(1234)
samples <- simulAr1Ar1(n=100, R=40, C=45, rho.r=0, rho.c=0.8)
stats <- list()
stats$cor.btw.rows <- c(apply(samples, 3, function(mat){
  apply(mat, 2, function(row.i){ # per column
    acf(row.i, lag.max=1, type="correlation", plot=FALSE)$acf[2]
})}))
stats$cor.btw.cols <- c(apply(samples, 3, function(mat){
  apply(mat, 1, function(col.j){ # per row
    acf(col.j, lag.max=1, type="correlation", plot=FALSE)$acf[2]
})}))
sapply(stats, summary)

## strong correlation between rows and between columns
set.seed(1234)
samples <- simulAr1Ar1(n=100, R=40, C=45, rho.r=0.8, rho.c=0.8)
stats <- list()
stats$cor.btw.rows <- c(apply(samples, 3, function(mat){
  apply(mat, 2, function(row.i){ # per column
    acf(row.i, lag.max=1, type="correlation", plot=FALSE)$acf[2]
})}))
stats$cor.btw.cols <- c(apply(samples, 3, function(mat){
  apply(mat, 1, function(col.j){ # per row
    acf(col.j, lag.max=1, type="correlation", plot=FALSE)$acf[2]
})}))
sapply(stats, summary)

## low correlation between rows and strong correlation between columns
set.seed(1234)
samples <- simulAr1Ar1(n=100, R=40, C=45, rho.r=0.2, rho.c=0.8)
stats <- list()
stats$cor.btw.rows <- c(apply(samples, 3, function(mat){
  apply(mat, 2, function(row.i){ # per column
    acf(row.i, lag.max=1, type="correlation", plot=FALSE)$acf[2]
})}))
stats$cor.btw.cols <- c(apply(samples, 3, function(mat){
  apply(mat, 1, function(col.j){ # per row
    acf(col.j, lag.max=1, type="correlation", plot=FALSE)$acf[2]
})}))
sapply(stats, summary)

## low correlation between rows and strong correlation between columns
## AND high error variance
set.seed(1234)
samples <- simulAr1Ar1(n=100, R=40, C=45, rho.r=0.2, rho.c=0.8,
                       sigma.X.2=1, sigma.e.2=200)
stats <- list()
stats$cor.btw.rows <- c(apply(samples, 3, function(mat){
  apply(mat, 2, function(row.i){ # per column
    acf(row.i, lag.max=1, type="correlation", plot=FALSE)$acf[2]
})}))
stats$cor.btw.cols <- c(apply(samples, 3, function(mat){
  apply(mat, 1, function(col.j){ # per row
    acf(col.j, lag.max=1, type="correlation", plot=FALSE)$acf[2]
})}))
sapply(stats, summary)

## End(Not run)

timflutre/rutilstimflutre documentation built on Feb. 7, 2024, 8:17 a.m.