Re: [igraph] errors using power.law.fit

[Gábor Csárdi]

Martin,

On Tue, Nov 16, 2010 at 9:58 AM, Martin Tomko <address@hidden> wrote:
> Dear all,
> I am running a batch process on a few hundreds graphs, and trying to get
> their power.law.fit exponents. Every now and then I get an error:
>
> Error in optim(start, f, method = method, hessian = TRUE, ...) :
>  non-finite finite-difference value [1]
>
>
> Which I assume arises from the use of power.law.fit - google seems to imply
> that optim() is related to statistical fitting.

you don't need google for this, the source code and documentation is all there.

Look at the source code of 'power.law.fit' and then you see it calls
'mle' from the 'stats4' package, which in turn calls 'optim'.  The
manual page of these functions explain that additional arguments to
'power.law.fit' are passed to 'mle' and then to 'optim', so you can
try another, more robust optimization method, e.g. BFGS easily:

power.law.fit(..., method="BFGS")

Best,
Gabor

> Now, it is possible that
> some of the graphs I have would not fit a power law well - is there a
> recommended way for checking whether it makes sense at all to try fitting
> one?
>
> Thanks,
> Martin
>
> _______________________________________________
> igraph-help mailing list
> address@hidden
> http://lists.nongnu.org/mailman/listinfo/igraph-help
>



-- 
Gabor Csardi <address@hidden>     UNIL DGM