MCLUST: finding the best number of Gaussian components to fit my data
0
0
Entering edit mode
@travascioandrea91-19699
Last seen 5.2 years ago

Hi everybody, my question is the following..

I have a sample of galaxy radial velocities in a galaxy cluster (unfortunately the size of this sample is N=18, I know..N<20 is not the best) and I wish to know what is the number of Gaussians which fit my data distribution in the best way [this can assume the values G=1:3]. Afterthat, I want to know what are the best Gaussian parameters. I expect G=3 as a best result of number of Gaussians to consider, but I need a number (I guess the log likelihood) that describes the significance of this case. Many use MCLUST (R package) for modeling data as a Gaussian finite mixture. I read that It allows to find the optimal number of components (through a clustering hierarchical approach) and the corresponding classi fication.

I tried to use the following pipelines:

1)Mclust with only these parameters...


> modClust = Mclust(dataset,G=1:3,modelsName="V")
fitting ...
|==============================================================| 100%
> summary(modClust)
%---------------------------------------------------- 
Gaussian finite mixture model fitted by EM algorithm 
%---------------------------------------------------- 
Mclust X (univariate normal) model with 1 component: 
 log.likelihood  n df      BIC      ICL
      -157.1966 18  2 -320.174 -320.174
Clustering table:
 1 
18

but it returns the best G value =1.. but I know it should be 3

2) I thought an alternative method could be to perform a FOR cycle in which I change the G value and I compare the log likelihood values..


Have you advices? What am I doing wrong? What am I not considering?

thanks in advance for the help,

Andrea

software error microarray probe annotation • 1.1k views
ADD COMMENT
0
Entering edit mode

I'm removing the "DESeq2" tag as I can't see any relevance to DESeq2.

ADD REPLY

Login before adding your answer.

Traffic: 972 users visited in the last hour
Help About
FAQ
Access RSS
API
Stats

Use of this site constitutes acceptance of our User Agreement and Privacy Policy.

Powered by the version 2.3.6