My Problem is the following. I have
Y ~ Normal(a,b)
X|Y=y ~Normal( f(y,j*))
where j* is to be choosen that will maximize the int(int(j*.x|y . f(x|y)).f(y)).
Therefore, I am using the Gauss-hermite procedure
a) to calculate the nodes
b) for those nodes evalue the optimal j* - using the optimization routine
c) then evalute conditional expectation for all the nodes
d) Then use the weights to compute the unconditinal expectation.
So, I am not sure if I can just straight away use the standard numerical integration package - because the function that I have to integrate over has to be endogeously determined.
I am sorry if I am not clear about things or being a little dense. Any advice/comments/suggestions to make this procedure faster is appreciated.
Thanks once again!