Support to efficiently cluster multivariate longitudinal data with sparse (very little observed points for each subject) and irregular (the observation occassions are not aligned across subjects) observations.
Longitudinal data with multiple outcomes is supported as well, where some of them could be pure noise (non-distinguishable).
Support the case with potentially unbalanced cluster size, i.e., the number of subjects in some clusters are way outnumbered by the others.
In sum, the package is capable of clustering sparse, irregular, unbalanced, and multivariate continuous longitudinal data
To install the package:
devtools::install_github("junyzhou10/clusterMLD")Use main function LongDataCluster(x, Y, id, ...) to cluster longitudinal data in long format. Parallel computing is supported by specifying parallel = TRUE.
DendroPlot(Cluster.object) yields corresponding dendrogram, where Cluster.object is the output from LongDataCluster()
MeanPlot(Cluster.object) yields corresponding mean curves of each detected cluster.
The author created an R shiny app for a better illustration/visualization of the package. Please refer to clusterMLD_ShinyApp for more details.
The App is published at https://junyzhou.shinyapps.io/clusterMLD_ShinyApp/, with two toy examples uploaded already. Please note that it will take more time to run parallel online than local.
The current version mainly support for the continuous outcomes. Algorithm for other outcome types, such as binary and time-to-event, are now under development.
Junyi Zhou, Ying Zhang & Wanzhu Tu (2022) clusterMLD: An Efficient Hierarchical Clustering Method for Multivariate Longitudinal Data, Journal of Computational and Graphical Statistics, DOI: 10.1080/10618600.2022.2149540