68
Let mi = E(wi ∣ N), when maximizing the utility function, straightforward alge-
bra shows that the optimal rule is
d,s = I{mi >c∕{k+ 1)).
(3.35)
Once we have a simulated sample (/jɪ, ðɪ),..., (∕¾v, J∙w) of δi) given the data.
We can estimate m⅛ with
1 M
⅛⅛∑M>tf>⅛
(3.36)
1 fc=l
and then make the decision diβ.