search for: 2_i

...modeling, if this is at all possible. For example we can take schools data by Gelman from R2WinBUGS package. As you can see bellow use of weights argument did not had influence on results. I do not know if my specification of weights i.e. 1 / sd^2 is ok. Under least squares one minimizes sum(e^2_i) or sum(w_i * e^2_i) with weighted LS. If I consider that \sigma_i represents uncertainty in my "data" then e'_i = e_i / \sigma_i and we minimize sum(e'^2_i) = sum((e_i / \sigma_i)^2) = sum(e_i * \sigma^-2_i). Therefore weights i.e. w_i are equal to 1 / \sigma^2_i. Can anyone...