function Z=xbivarn(ux, sigx, uy, sigy, rho, samples, seed) 
% xbivarn(ux, sigx, uy, sigy, rho, samples, seed) Bivariate normal 
%    experiment.  Generates an "unseen" normal random variable X with mean 
%    Then from X generates a "measured" normal random variable Y, and uses a 
%    linear estimation to determine X from Y.  The output is a matrix, in 
%    which each row represents an experimental parameter 
% 
%    row   description 
%    1     "unseen" variable x 
%    2     "measured" variable y 
%    3     x' : estimated x values from y values 
%    4     y' : estimated y values from x values 
%    5     error between x and x' 
%    6     error between y and y' 
if sigx<=0 | sigy<=0 
   error('sigma squared cannot be <=0') 
end 
if samples<=0 
   error('invalid value for samples') 
end 
G=rho*sqrt(sigy/sigx); 
npdf1=dfnorm(0, sigx); 
npdf2=dfnorm(0, sigy*(1-rho^2)); 
N1=rvgen(npdf1, 1, samples, seed); 
N2=rvgen(npdf1, 1, samples, seed^2); 
X=N1+ux; 
Y=N1.*G+N2+uy; 
X1=rho.*(Y-uy)./sqrt(sigy).*sqrt(sigx)+ux; 
Y1=(1/rho).*(X-ux)./sqrt(sigx).*sqrt(sigy)+uy; 
EX=(X-X1); 
EY=(Y-Y1); 
Z=[X; Y; X1; Y1; EX; EY];