Re: Fitting with leasqr and uncertainties

[Mike Miller]

[Moving this followup to the maintainers list.]

On Tue, Jan 21, 2014 at 11:42 AM, CdeMills wrote:
> st.michal88 wrote
>>   what do I fit with:
>> Function is more or less: y(t)=param1*exp(-(t-param2))+param3
>
> Hello,
>
> Carnë has created a new entry. For now:
> 1) cd to some place inside your octave dir
> 2) type 'hg clone hg.code.sf.net/p/octave/prony"
> 3) this will create a new dir whose name is "prony". Add 'prony/inst' to
> your PATH
> [...]
> Try, check, and please provide feedback.

Hi Pascal, I'm not familiar with this algorithm yet, but the name
reminds me of some items that have come up before.

Can you comment on how this differs from the Matlab function prony in
the signal processing toolbox [1]? There is no prony function in the
signal package yet, but there should be eventually. Is there any
overlap with what you have created here? Is this exactly what prony
should be, or similar and can be adapted to the Matlab calling
interface?

Also see bug #38063 [2] for some discussion on the expfit function in
the optim package, where the Prony method was discussed also. Is there
some overlap with that function as well or bits that can be borrowed?

[1] http://www.mathworks.com/help/signal/ref/prony.html
[2] http://savannah.gnu.org/bugs/?38063

Thanks,

-- 
mike