trajeR, an R package for cluster analysis of time series

Abstract

We present the R package trajeR which provides all necessary tools to calibrate generalized finite mixture models, plot the results graphically and test the model adequacy.
First, we give an overview of the generalized finite mixture model for clustering time series and describe the core function trajeR of the package trajeR.
We give a step-by-step guideline on how to use the different available functions on simulated data sets with the three main underlying data distributions that appear in practice, the binary logit distribution, the (censored) normal distribution
and the zero inflated Poisson distribution.
Even though the estimation results are usually very similar, the trajeR package supports the two main estimations methods which are direct optimization of the derivative of the loglikelihood
and the Expected Maximization Algorithm. It also allows to compute group membership probabilities for all subjects in the data set, to exhibit profiles for the clusters found in the analysis and provides
a whole bunch of model adequacy criteria, together with the main model selection criteria.

Keywords: R package, generalized finite mixture model, fuzzy logic cluster analysis of trajectories

JEL Classification: C38, C88