Skip to content

Simulating correlations with varying sample sizes

4 messages · Jorge Ivan Velez, Holger Steinmetz, Dennis Murphy

Original
Holger Steinmetz
# 
Hi there,

I would like to draw 10 correlations from a bivariate population - but every
draw should be done with a different sample size. I thought I could to this
with a loop:

r=numeric(10) #Goal vector 
N = c(1000,100,80,250,125,375,90,211,160,540) #Sample size vector
for(i in 1:10) {
    data <- mvrnorm(n=N,mu=c(0,0),Sigma=matrix(c(1,.3,.3,1),2)) 
    r[i] <- cor(data[,1],data[,2])
    }

Goal: The 10 correlations shall be contained in the r-vector.
However, this does not work. I get an error that "arguments do not match"

Has anybody an idea?

Best,
Holger



--
View this message in context: http://r.789695.n4.nabble.com/Simulating-correlations-with-varying-sample-sizes-tp3526231p3526231.html
Sent from the R help mailing list archive at Nabble.com.
Dennis Murphy
# 
Hi:

A more 'R-ish' way to do this would be to use sapply() in place of the
loop as follows:

N = c(1000,100,80,250,125,375,90,211,160,540)
corrfun <- function(n) {
   require(MASS)
   dat <- mvrnorm(n=n, mu=c(0,0), Sigma=matrix(c(1,.3,.3,1), 2))
   r <- cor(dat)[1, 2]
   }
sapply(N, corrfun)

# My result:
[1] 0.29095517 0.42442863 0.04917587 0.28322478 0.42035862 0.30463885
 [7] 0.28411963 0.35413437 0.32901814 0.26318604

In this case, sapply() applies corrfun() to each element of the vector
N and outputs a numeric vector, so there is no need to initialize a
vector in advance of a loop. The timings are about the same (0.45 sec
for 100 iterations of sapply() vs. 0.47 sec for 100 iterations of the
corrected loop), but the code is clearer (at least to R-helpers :) and
more compact.

HTH,
Dennis

On Mon, May 16, 2011 at 6:42 AM, Holger Steinmetz
<Holger.steinmetz at web.de> wrote: