Re: statistics package more suitable for nlinfit et al.?

[Julien Bect]

Le 19/08/2015 12:35, Olaf Till a écrit :
> After some modifications and fixes to docs and demo, and a
> modification in the way weighted residuals are computed, nlinfit.m is
> now ready for inclusion into Octave Forge, as are statset.m,
> statget.m, and __all_stat_opts__.m.
>
> Do we want the degree in Matlab compatibility to have these functions
> in the statistics package, which corresponds to the 'correct' Matlab
> package? It would make the statistics package depend on the optim
> package, since nlinfit wraps nonlin_curvefit and curvefit_stat.
>
> Olaf

Is there a general policy about such things in the Octave Forge project? 
If not, should we adopt one?

Is this a goal of the Octave Forge project to provide one-by-one 
replacement for some (if not all) Matlab toolboxes ? Or is it enough to 
point to the right direction when a user asks for a function, like this:

>> nlinfit
warning: the 'nlinfit' function belongs to the statistics package from 
Octave
Forge which you have installed but not loaded.  To load the package, run
'pkg load statistics' from the Octave prompt.

?

Is there somewhere a list of the Octave Forge packages that aim at 
providing a compatible replacement for a corresponding Matlab toolbox?

For instance, how can I know for the Octave Forge website whether the 
"statistics" package aims at being a compatible replacement for the 
Mathworks' "Statistics Toolbox"? And what about control, signal, 
symbolic... ?

Just my two cents :-)