Skip to content

mclust for time series

2 messages · atesh koul, Ranjan Maitra

Original
atesh koul
# 
Hi all,

I am new to using model based clustering. I am using the clustering to find
how many clusters could be there in the data and the assignment of trials
to these clusters.

The data are hand kinematics data- 12 variables at 10 discrete time points
for 17 subjects. I have previously seen use of mclust for time series.
However, I am not sure how it would be applied in terms of the formula. I
have seen couple of papers that use it from a theoretical point of view
(but not as a command in R). I would like to know if mclust can be used for
multivariate time series data.

I would really appreciate any help with this.

Thanks in anticipation,


Best,
Atesh
Ranjan Maitra
#
On Mon, 3 Nov 2014 13:51:52 +0100 atesh koul <atesh.koul at gmail.com> wrote:

            

      
      
      
    

        
Atesh,

I am not sure how useful it will be to cluster 17 12-variate observations (subjects) using (presumably) Gaussian model-based clustering, but you may look at the following paper:

Chen, W.-C. and Maitra, R. (2011). Model-Based Clustering of Regression Time Series Data via APECM
? An AECM Algorithm Sung to An Even Faster Beat. Statistical Analysis and Data Mining, 4:567-78. DOI
information: 10.1002/sam.10143.

There is R code available (not R package), but again, it is not clear to me how useful this is going to be unless there is a structure in the 12 variables you are willing to give in (of course, unless I have misunderstood the context of your question). 

Best wishes,
Ranjan