Discover MakerZone

MATLAB and Simulink resources for Arduino, LEGO, and Raspberry Pi

Learn more

Discover what MATLAB® can do for your career.

Opportunities for recent engineering grads.

Apply Today

Thread Subject:
mle and fminsearch

Subject: mle and fminsearch

From: lucie

Date: 20 Jul, 2011 16:44:09

Message: 1 of 4

I want to minimize the following log-likelihood function which is the log likelihood function for normal half-normal distribution.
function L = (n/2)*log(0.5*pi*sig^2)-sum(log(normcdf((y-x1*B)*lam/sig)))+0.5*sig^1/2*(y-x1*B)'*(y-x1*B);
 the purpose is to find the Bs, and the sigma and lamda that minimize the L, while the y is n*1 vector, and x1 is n*k matrix, where n=6000, k=30.
So there is 30 parameter estimation for B, and one for sigma and lamda respectively.
I try to estimate the function using fminsearch, or fminunc but no results returned, I wonder if I used the right technique to solve this kind of problem?
It will be appreciated if there is any suggestion provided.

Best,
Lucy

Subject: mle and fminsearch

From: Matt J

Date: 20 Jul, 2011 17:47:09

Message: 2 of 4

"lucie " <luciedeng@163.com> wrote in message <j070kp$oof$1@newscl01ah.mathworks.com>...
>
> So there is 30 parameter estimation for B, and one for sigma and lamda respectively.
> I try to estimate the function using fminsearch, or fminunc but no results returned, I wonder if I used the right technique to solve this kind of problem?
=====================

FMINSEARCH probably isn't a good choice considering the large number of variables.

Other than that, elaborate on "no results returned". For example, the following statement returns no results either, but that shouldn't be a surprise.

>>exp(1:5);

Subject: mle and fminsearch

From: lucie

Date: 20 Jul, 2011 19:37:09

Message: 3 of 4

"Matt J" wrote in message <j074at$6rd$1@newscl01ah.mathworks.com>...
> "lucie " <luciedeng@163.com> wrote in message <j070kp$oof$1@newscl01ah.mathworks.com>...
> >
> > So there is 30 parameter estimation for B, and one for sigma and lamda respectively.
> > I try to estimate the function using fminsearch, or fminunc but no results returned, I wonder if I used the right technique to solve this kind of problem?
> =====================
>
> FMINSEARCH probably isn't a good choice considering the large number of variables.
>
> Other than that, elaborate on "no results returned". For example, the following statement returns no results either, but that shouldn't be a surprise.
>
> >>exp(1:5);

Thanks, but what technique I can use to solve this kind of problem? I checked the Maximum Likelihood estimation(MLE), it only provides estimation on the parameters such as sigma, roh of a vector, but in the situation y=Xb+u, if the distribution of the disturbance 'u' is known, say in my case, normal half normal distribution, and I know its log likelihood function, what is the typical way to derive the estimation on the bs, as well as the other parameters say sigma and lamda, etc.

hope to hear from you soon,
Regards,
Lucie
    

Subject: mle and fminsearch

From: Matt J

Date: 20 Jul, 2011 21:41:10

Message: 4 of 4

"lucie " <luciedeng@163.com> wrote in message <j07ap5$qfd$1@newscl01ah.mathworks.com>...
>
> Thanks, but what technique I can use to solve this kind of problem? I checked the Maximum Likelihood estimation(MLE), it only provides estimation on the parameters such as sigma, roh of a vector, but in the situation y=Xb+u, if the distribution of the disturbance 'u' is known, say in my case, normal half normal distribution, and I know its log likelihood function, what is the typical way to derive the estimation on the bs, as well as the other parameters say sigma and lamda, etc.
================

I only said FMINSEARCH was a bad choice. I never said FMINUNC was bad. Assuming there are no constraints you'd like to put on your parameters, it seems appropriate enough to try.

Tags for this Thread

No tags are associated with this thread.

What are tags?

A tag is like a keyword or category label associated with each thread. Tags make it easier for you to find threads of interest.

Anyone can tag a thread. Tags are public and visible to everyone.

Contact us