On Sep 8, 2010, at 10:34 AM, NatsumiYotsumoto wrote:
Dear all.
I'm using igraph package, and do a research about network analysis.
With power.law.fit from igraph package, it seems that we can fit a
power law
distribution to some data.
But, I want to know how to judge whether the network distribution
follows a
power law or not.
In order to determine whether something is from distribution A or not-
A, one needs to have a sensible way of characterizing or considering
what would be in the range of distributions in the not-A.
Unfortunately for your question, the range of possible distributions
is infinite. That means it would always be possible to have a better
fitting distribution than what ever is distribution A. If you have
alternatives to the power-law that you want to put to the test, then
now is the time to offer them.
My guess is that you do not, so I will offer alternatives:
Alt A:
a) read the citations in the email you cited, especially Newman then ...
b) set up a histogram of your data using hist with logarithmic or
geometric progression of the breaks argument.
c) as a check on you exponent estimate, calculate alpha and se(alpha)
as on pg 4-5 of that citation.
Alt B:
require(sos)
???fitting pareto
???fitting power network # and proceed from there
--
David.
Does anyone know the way to do this?
Thanks for any help.
Daigo
p.s.
Also, I tried several ways such as
http://www.mail-archive.com/r-h...@stat.math.ethz.ch/msg62520.html
and I got results like this:
Profiling...
2.5 % 97.5 %
2.393297 2.412650
What do these suggest?
please tell me about this if someone knows.
--
David Winsemius, MD
West Hartford, CT
__
R-help@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.
