Take the 2-minute tour ×
Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. It's 100% free, no registration required.

Suppose that we have two independent normal distributions $\mathcal{N}_{1}(0,s)$, $\mathcal{N}_{2}(0,t)$ what is the distribution of empirical covariance (or empirical correlation if this make my question simpler ) between two samples of (fixed) size $k$ ( observations are indexed by ${t}$ ) sampled from $\mathcal{N}_{1}(0,s)$, $\mathcal{N}_{2}(0,t)$ ?
Simply approach involving convolution seems to doesn't work for me well, because $2k-1$ dimensional integral appears with which I can't deal.

I've made some simulations for $k=10,9,8,...,2$ we have densities :

enter image description here

here is simple R code :

cov_<-function(k){
   cov(rnorm(k),rnorm(k))
}

sim<-function(n,ile){
   wyn<-numeric(n)
   for(i in 1:n){
      wyn[i]=cov_(ile)
   }
   wyn
}



plot(density(sim(10000,10)))

for(j in 9:2){
    lines(density(sim(10000,j)),col=j-1)
}
share|improve this question
add comment

1 Answer

The empirical covariance converges to the theoretical covariance (in your case, zero) at the rate $1/\sqrt{k}$ when $k\to\infty$ but it has no simple distribution for fixed finite $k$. See the first pages of this paper to get an idea of the situation.

share|improve this answer
add comment

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.