Arun Kumar Saha <arun.kumar.saha <at> gmail.com> writes:
> hi all,
> suppose I have:
> r[i] ~ N(0, h[i])
> h[i] = a + b*r[i-1] + c*h[i-1] for all i=2..........n
> I want to get estimates of a, b, c by mle.
> Can you tell me how to do that?
(1) can you convince us this isn't a homework problem?
(a little bit more context would be nice in any case)
(2) are both r and h observed? if so, then
your (negative log-)likelihood function can use
something along the lines of
where you fill in the predicted mean and std. dev. of each.