1
$\begingroup$

FindDistributionParameters (and EstimatedDistribution) can use "MethodOfMoments" as its ParameterEstimator, but as far as I can tell it still requires you to give it the full observed data as its first argument. If I have the moments, but not the full data, is there any way to just pass the observed moments directly to FindDistributionParameters? Or do I have to write down the system of equations for the moments myself and then use FindRoot?

Improve this question
$\endgroup$
4
  • 2
    $\begingroup$ You will have to do this yourself but depending on the distribution it shouldn't be difficult. If you need help with a particular example you can always add it to your question. $\endgroup$
    – Andy Ross
    Dec 5, 2014 at 5:24
  • $\begingroup$ Or -- a somewhat roundabout approach -- you could generate lots of data using the parameters you have and then use FindDistributedParameters on the data you have generated. $\endgroup$
    – bill s
    Dec 5, 2014 at 15:23
  • $\begingroup$ @bills however if the moments they have are the parameters they wouldn't need to estimate. I think they have a situation where the parameters of interest are some function of the moments $\endgroup$
    – Andy Ross
    Dec 5, 2014 at 18:49
  • $\begingroup$ @AndyRoss That's correct in my case -- if I knew how to generate the data I'd be done! It was just a bit annoying to write it myself, because writing the moments in terms of the parameters involved special functions, and I had to play with the form a bit to avoid over/underflows. I guess that means that I probably wouldn't have been able to use the default ParameterEstimator in any case though. $\endgroup$
    – Daniel Weissman
    Dec 5, 2014 at 19:25

1 Answer 1

Reset to default
1
$\begingroup$

There is an add-on for Mathematica, called MathStatica, which does exactly what you want, if I understood you right.

Find an example on www.mathstatica.com/examples and click on the left side on the item "Person Fitting." Ths is an excellent add-on which fits seamless into mma and extends its statistical capabilities.

Best regards Volker

Improve this answer
$\endgroup$
1
  • $\begingroup$ I was looking for something that would try to fit an arbitrary parametric form with an arbitrary number of free parameters (matching the number of moments). $\endgroup$
    – Daniel Weissman
    Dec 10, 2014 at 17:44

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

Not the answer you're looking for? Browse other questions tagged or ask your own question.