muhammad mohsin <mohsinshahid <at> yahoo.com> writes:
> Hope you will be fine. I am student of Ph.D and doing '
> some work on distribution.
> I developed a new distribution and having some problems in estimating their
> parameters by MLE. I used R-program and? used "maxLik" function
(maxLik: A
> Package for Maximum Likelihood Estimation in R)
> But there is some problem, it is
> not estimated the parameters properly.
> I also write an e-mail to the author of
> this paper but he could not solve my problem.
> His function works well for simple
> and known distribution but does?not work for a new function.??
> Can?anybody spare
> some time for me?? I really need your help.
> Please inform me so that I can send
> you the material.
> Waiting for a quick reply
> Best Regards,
>
> ?Muhammad Mohsin
Dear Mr Mohsin,
You are free to post a (self-contained/reproducible and preferably
small/minimal) example here and see if it interests anyone sufficiently
for them to volunteer time to see if they can find the problem. With respect,
though, if you are a PhD student in statistics then this is part of your
training, and it should really fall to you, or to your supervisor or other
people at your institution, to work out how to solve it. You are (much)
more likely to get useful help from this group if you can narrow
your problem down to a specific
point, and if you can indicate what steps you have tried to take to
solve your problem for yourself. Maximum likelihood estimation is
in general a challenging computational problem -- just because a
general-purpose function or package exists doesn't mean it can solve
all problems easily. You may have to work harder to understand
the particular structure of your optimization problem and what
methods will work for it.
For a start, you might try other optimization algorithms (see e.g.
the 'optimx' package, which may be on R-forge rather than CRAN
[I don't remember], as well as the Optimization task view on CRAN).
good luck
Ben Bolker