MATLAB Answers


Zero-Inflated and Hurdle Models in Matlab (statistical distribution fitting)

Héctor de Santos さんによって質問されました 2019 年 1 月 20 日
最新アクティビティ Jeff Miller
さんによって コメントされました 2019 年 4 月 29 日
I have a very zero-inflated dataset and I am trying to find it in an statistical distribution. I tried several distributions in Matlab (distribution Fitter App) and no one works.
Now I want to to use a zero-inflated or hurdle model, however I do not find any reference nor example in matlab. How could I fit my data to a hurdle distribution in matlab?, in google the only reference I found was PSCL package for R software, but I would like to continue my work in Matlab.Is there something similar?
Thank you .

  0 件のコメント

サインイン to comment.

1 件の回答

回答者: Jeff Miller
2019 年 1 月 20 日

You might be able to fit the distributions you have in mind using Cupid. For example, the following commands would define and plot zero-inflated Poisson and Binomial distributions (at least as I understand the terminology):
zipois = Mixture(.2,ConstantD(0),Poisson(4))
zibin = Mixture(.2,ConstantD(0),Binomial(4,.8))
Distribution parameters and mixture probabilities can be tweaked by hand or estimated directly from data (e.g., maximum likelihood). For example,
fakedata = zibin.Random(2000,1);
figure; histogram(fakedata);
zibin.EstML(fakedata,'rffr') % rffr means: Adjust the real-valued mixture probability,
% hold fixed the constant value and the binomial N
% adjust the real-valued binomial probability

  2 件のコメント

Jeff Miller
2019 年 4 月 29 日
I am not sure what is going on with the link in the answer above. This is the correct link:

サインイン to comment.

Translated by