conflicting results of lognormal fit

[likec...@gmail.com]

Hi,

After using powerlaw_compare, I found my upper tail data better fit to lognormal. Then I want to give a best estimation of the lognormal parameter. But I found there are two different ways of calculating lognormal fit. One is using lognormal.mu after using powerlaw.Fit, the other is using powerlaw.distribution_fit directly. In the latter case I set the xmin according to calculation of best xmin in powerlaw.Fit. They returned different results and I have no idea which one is right.

BTW, my ultimate goal is to give an estimation of lognormal parameter at every xmin, just as we did for powerlaw fit. And then through comparing D-statistics to choose a best xmin for lognormal fit. I know it should be something like another package "lognormal fit", I just wonder if there're any possibilities of doing this within powerlaw package. And if not, the previous question is important to me in that I need to know the way it is calculated.

Thank you very much!

[Jeff Alstott]

On Fri, Mar 17, 2017 at 9:26 PM, <likec...@gmail.com> wrote:

After using powerlaw_compare, I found my upper tail data better fit to lognormal. Then I want to give a best estimation of the lognormal parameter. But I found there are two different ways of calculating lognormal fit. One is using lognormal.mu after using powerlaw.Fit, the other is using powerlaw.distribution_fit directly. In the latter case I set the xmin according to calculation of best xmin in powerlaw.Fit. They returned different results and I have no idea which one is right.

That does not sound good. Can you give a minimal working example as an issue on Github?

 

BTW, my ultimate goal is to give an estimation of lognormal parameter at every xmin, just as we did for powerlaw fit. And then through comparing D-statistics to choose a best xmin for lognormal fit. I know it should be something like another package "lognormal fit", I just wonder if there're any possibilities of doing this within powerlaw package. And if not, the previous question is important to me in that I need to know the way it is calculated.

That's definitely a thing that could be created. You could fork the code and modify the behavior of `find_xmin` to include an option for what distribution you're using as the reference (i.e. Power_Law vs. Lognormal)

https://github.com/jeffalstott/powerlaw/blob/master/powerlaw.py#L171

[李可纯]

Hi Jeff,

Thank you so much for your quick reply. I posted an issue on Github now (#38). The example is almost the same as the screenshot in my last email. I guess it's due to my misunderstanding but I just cannot figure out why.

And your guidance of the code seems terrific. I'll try:D

--
You received this message because you are subscribed to a topic in the Google Groups "powerlaw-general" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/powerlaw-general/h7hzFfLeNVM/unsubscribe.
To unsubscribe from this group and all its topics, send an email to powerlaw-general+unsubscribe@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[Jeff Alstott]

(The answer was: "distribution_fit old code that is unadvertised and no one should be using and should be removed from future versions")

--
You received this message because you are subscribed to the Google Groups "powerlaw-general" group.
To unsubscribe from this group and stop receiving emails from it, send an email to powerlaw-general+unsubscribe@googlegroups.com.