I would just use the mclust R-package that can handle 1D data well. You can include your data as an argument in the Mclust-function and then easily create plots with your mclust-object -> plot(my_mclustobject). Further statistics can be accessed without problem: e.g. my_mclustobject$uncertainty or my_mclustobject$classification

When you use Mclust, you can also enter the number of clusters you want to check as an argument (e.g. G=1:15)

Here is a full documentation:  ftp://gnu.cs.pu.edu.tw/network/CRAN/web/packages/mclust/vignettes/mclust.pdf

For 1D-data two possible models are available: univariate distribution with either equal or variable volume.

Thank You for Your answer! Mostly I use python or matlab, I'm not used to R. But mclust was a good keyword to find the scikit-package with GMM for python, which could to be exactly what I need. I will dig the next hours/days inside. I hope to be able to give some positive feedback next time :)

http://scikit-learn.org/stable/modules/mixture.html

Ok, I think in Python you can find also many examples with code: http://www.astroml.org/book_figures/chapter4/fig_GMM_1D.html

Good luck!

That is the only example could find for 1d, and the problem is, that the .eval-method seems to be removed from actual version of sklearn. I try to find an equal method for the currently implemented GMM or GaussianMixture function, I tried all method listed there http://scikit-learn.org/stable/modules/generated/sklearn.mixture.GaussianMixture.html, but was not successful.

With GMM it is easy to achieve the means and covariances, so it would be already enough, but it would be nice to use the package function, without building own GMM function for evaluation and plotting

Try the R-package; it's quite easy. You just need to import your data e.g. as CSV, then use the package. I highly recommend Rstuido; it's a lot easier to start with R. Then just follow the code in the vignette of the package. If you have any questions about that just write me a message. As output you get a mclust-object then you can extract all information you want. Maybe the right time to start using R ;-)

I got very nice results with the package sklearn now. Exactly what I need, very fast and stable! I can recommend this package! ;)

It is better for me to stay with python. So I can integrate my scripts in each other. Nevertheless thank You so much for Your answers, especially for the keyword mclust!

I just get the same problem and wondering how did you solve the problem with sklearn?

Hi Junjie,

I assume you have already checked the link from Adrian?

http://www.astroml.org/book_figures/chapter4/fig_GMM_1D.html

The most important things I used were:

from sklearn import mixture

mixture.GMM(n_components=i+1,n_iter=500).fit(hut)

"hut" is just the 1d-array with function values, not the density function

Thanks for your help! I want to go with that example but failed.

I just checked out this page https://github.com/scikit-learn/scikit-learn/issues/4877, and changed the following code:

s = x[:,np.newaxis] logprob, responsibilities = M_best.score_samples(s)

Then the example works, but it seems that the class GMM is deprecated.