Prev: how to plot time in the (hr/min/sec/millisec) format
Next: What is fastest video format for mmreader?
From: Peter Perkins on 4 Nov 2009 08:36
Yudha wrote:

> The way I use mle is like this :
> [phatl,pcil]=mle(t,'logpdf',lpdf,'start',[5/3,1.4],'lower',[5/3,-inf],'upper',[2,inf]);

You are giving a starting value for the first parameter right at the lower limit. Don't do that.

> But now the 0 values is the likelihood itself after we calculate using estimated parameter values given by mle. While the log_likelihood is giving values in the order of -2.10^4

Yes, as I said, that's why one works with the LOG-likelihood. What's exp(-2e4)? A number smaller than anything you can represent in double precision floating point. Why do you need to compute the unlogged likelihood?

> What actually is the numerical scheme that Matlab use for mle ?, is it Newton-Raphson or else ?. Thanks

As the help says:
>> help mle
MLE Maximum likelihood estimation.
[snip]
'optimfun' A string, either 'fminsearch' or 'fmincon', naming the
optimization function to be used in maximizing the
likelihood. Default is 'fminsearch'. You may only
specify 'fmincon' if the Optimization Toolbox is
available.

 | 
Pages: 1
Prev: how to plot time in the (hr/min/sec/millisec) format
Next: What is fastest video format for mmreader?