Any ideas how to reduce run time with mvncdf?

Hi all,

My code is the following:

q=1;

while q<=3

[w x y1 z]=ndgrid(A,A,A,A);

m=[2.326 1.466 0.9539];

parfor i=1:length(w(:))

fp(i)=mvncdf([w(i) x(i) y1(i) z(i)],[inf inf inf inf],zeros(1,n),[1 0.3 0.3 0.3; 0.3 1 0.3 0.3; 0.3 0.3 1 0.3; 0.3 0.3 0.3 1]);

tp(i)=mvncdf([w(i) x(i) y1(i) z(i)],[inf inf inf inf],repmat(m(q),[1 n]),[1 0.3 0.3 0.3; 0.3 1 0.3 0.3; 0.3 0.3 1 0.3; 0.3 0.3 0.3 1]);

oc(i)=(N*C(1))+C(2)*((1-normcdf(w(i),m(q),1))*pt+(1-normcdf(w(i)))*(1-pt))+C(3)*((1-normcdf(x(i),m(q),1))*pt+(1-normcdf(x(i)))*(1-pt))+C(4)*((1-normcdf(y1(i),m(q),1))*pt+(1-normcdf(y1(i)))*(1-pt));

end

curve=sortrows([fp' tp' oc'],1);

rcrv=rocc(curve(:,1),curve(:,2));

bc=(1-rcrv(:,2))+rcrv(:,1);

tc=g*curve(:,3)+(1-g)*bc;

cbn{q}=[rcrv(:,1) rcrv(:,2) curve(:,3) bc tc];

q=q+1;

end

I have matlabpool open in the local config (so that's not the issue). The issue here is that for this many elements, the computation is taking too long. I only need approximately five significant figures. I have to run scripts like this multiple times and my computer resources are modest. Does anyone have any thoughts on increasing the computability of this code? Thank you.

-Alex