Under the assumption that the data have a joint normal distribution, show that n 1/2 k

r,s,t,u has a limiting covariance matrix given by n cov (k i,j,k,l

, k r,s,t,u) → κ

i,rκ

j,sκ

k,tκ

l,u

[4!].

Hence show that n 1/2 pr̅4 has a limiting normal distribution with mean zero and variance 8p 2

+ 16p.