Re: [igraph] power law

[Gang Su]

Hi Thiago,

Here's a quick wiki:
http://en.wikipedia.org/wiki/Goodness_of_fit

If you have n bins of degrees, your goodness of fit will have n-1 
degrees of freedom on chi-square distribution.
What you want to do, is you will have an alpha after the fitting, so 
your pwr law dist will look like

f(x) = x to the pwr of (-alpha)

you plug in x = 1, 2, 3, 4, 5, 6... and you will have f(x), you multiply 
f(x) with the total number of degrees (2*num of edges), you may round
up or not, then that's the expected values

your actual degree counts will be observed

look up
chisq.test and ks.test in R,

plug in the obs and exp
then you will have a p-value

you can also measure by residuals, or other statistics

Gang





Thiago M. Venancio wrote:
> Hi Gang,
>
> Thanks for your reply.
> Could you possibly give more details on the goodness of fit test ?
>
> Best.
>
> Thiago
>
> On Mon, Jul 7, 2008 at 11:34 AM, Gang Su <address@hidden> wrote:
>   
> > Hi Thiago,
> >
> > I am not a developer, i don't think there's a straight forward way to
> > measure goodness of fit.
> > But you can always do the r squared or goodness of fit test on your own,
> > just take the observed
> > degree values and expected degree values, and plugin the goodness of fit
> > test.
> >
> > Gang
> >
> > Thiago M. Venancio wrote:
> >     
> > > Hi all,
> > >
> > > Given a degree distribution of a graph, is there any straightforward
> > > way to assess the likelihood/significance of the fitness using iGraph
> > > (e.g compare with other non power law distributions ?
> > >
> > > Thanks in advance.
> > >
> > > Thiagp
> > >
> > >
> > > _______________________________________________
> > > igraph-help mailing list
> > > address@hidden
> > > http://lists.nongnu.org/mailman/listinfo/igraph-help
> > >
> > >
> > >
> > >       
> >
> > _______________________________________________
> > igraph-help mailing list
> > address@hidden
> > http://lists.nongnu.org/mailman/listinfo/igraph-help
> >
> >     
>
>
> _______________________________________________
> igraph-help mailing list
> address@hidden
> http://lists.nongnu.org/mailman/listinfo/igraph-help
>
>
>