Alex
You can not use your approach as you can consider the distributions of N and X are independent only when you take them conditional on theta.
To see this you need to follow through the derivation of E(S) in terms of expectations of individual random variables X and N. You will see at the point when you invoke independence of X and N you need to assume them being conditional on theta. Otherwise your workings will be wrong.
Last edited: Apr 15, 2013