function val=normcau(th,y)
% 
% normal/cauchy model
% y is N(th, sig)
% th is Cauchy(m,tau)

sig=10; m=100; tau=6.75;

like=-.5*log(2*pi)-log(sig)-1/(2*sig^2)*(y-th).^2;
prior=-log(pi*tau)-log(1+(th-m).^2/tau^2);
val=like+prior;