1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
|
function y = MultiNormalDraw(n,u,K)
% x = MultiNormalDraw(n,u,K)
%
% Make n multivariate normal draws with mean u and covariance matrix K.
% Each draw is in a single column of y, which has n columns.
%
% The routine operates by computing the appropriate linear transformation
% of a N(0,I) multivariate normal draw. This transformation is given by
% y= C'x + u where K = C'C. This works because the covariance of a
% distribution y = Cx is in general given by Ky = C Kx C'. In our case
% Kx= I so Ky = C'C = K.
%
% K = 0 is handled as a special case
% 12/31/93 dhb Added this line.
% 7/24/04 awi Cosmetic.
[m,null] = size(u);
x = zeros(m,n);
if (MatMax(K) == 0)
C = zeros(m,m);
else
C = chol(K);
end
for i = 1:n
x(:,i) = NormalDraw(m,0,1);
y(:,i) = C'*x(:,i) + u;
end
|
