There are some packages to fitting power law to empirical date in "R", "matlab", "python" Here with maximum likelihood method As it said here. Does it exist similar package in Mathematica?
$\begingroup$
$\endgroup$
Start here to see what statistics functionality is available:
- http://reference.wolfram.com/language/guide/Statistics.html
- http://reference.wolfram.com/language/guide/ProbabilityAndStatistics.html
Fitting is done with EstimatedDistribution
or FindDistributionParameters
. You want to fit a ParetoDistribution
for continuous data or a ZipfDistribution
for discrete data.
There is no need for a separate package.
-
1$\begingroup$ Also, ParetoDistribution cross-references with PowerDistribution $\endgroup$ – Bob Hanlon Jun 11 '16 at 13:53
-
$\begingroup$ @BobHanlon thanks, Instead of data (because of their huge size) I saved distribution of data : instead of {32,21,48,52,32,32,32,32,21,48,21} I have {{32,5},{21,3},{48,2},{52,1}} , the repetition of each data, How can find parameter of Power Distribution with maximum likelihood method of that? $\endgroup$ – jack cilba Jun 13 '16 at 11:40
-
1$\begingroup$ @jackcilba You need
ParetoDistribtion
orZipfDistribution
, but not PowerDistribution, to do what is described in the paper you linked. To choose a fitting method, read the documentation ofEstimatedDistribution
that I already linked, especially theParameterEstimator
option. Be sure to check under "Details" when reading ay documentation page. $\endgroup$ – Szabolcs Jun 13 '16 at 12:43 -
3$\begingroup$ @jackcilba - for additional questions post a new question rather than asking in a comment. $\endgroup$ – Bob Hanlon Jun 13 '16 at 15:11