Abstract
In this article, a new method is proposed for clustering longitudinal curves. In the proposed method, clusters of mean functions are identified through a weighted concave pairwise fusion method. The EM algorithm and the alternating direction method of multipliers algorithm are combined to estimate the group structure, mean functions and principal components simultaneously. The proposed method also allows to incorporate the prior neighborhood information to have more meaningful groups by adding pairwise weights in the pairwise penalties. In the simulation study, the performance of the proposed method is compared to some existing clustering methods in terms of the accuracy for estimating the number of subgroups and mean functions. The results suggest that ignoring the covariance structure will have a great effect on the performance of estimating the number of groups and estimating accuracy. The effect of including pairwise weights is also explored in a spatial lattice setting to take into consideration of the spatial information. The results show that incorporating spatial weights will improve the performance. A real example is used to illustrate the proposed method.
Similar content being viewed by others
References
Basu S, Banerjee A, Mooney R.J (2004) Active semi-supervision for pairwise constrained clustering. In: Proceedings of the 2004 SIAM international conference on data mining. SIAM, pp. 333–344
Bouveyron C, Côme E, Jacques J (2015) The discriminative functional mixture model for a comparative analysis of bike sharing systems. Ann Appl Stat 9(4):1726–1760
Bouveyron C, Jacques J (2011) Model-based clustering of time series in group-specific functional subspaces. Adv Data Anal Classif 5(4):281–300
Boyd S, Parikh N, Chu E, Peleato B, Eckstein J (2011) Distributed optimization and statistical learning via the alternating direction method of multipliers. Found Trends Mach Learn 3(1):1–122
Chi EC, Lange K (2015) Splitting methods for convex clustering. J Comput Graph Stat 24(4):994–1013
Chiou JM, Li PL (2007) Functional clustering and identifying substructures of longitudinal data. J R Stat Soc Ser B (Stat Methodol) 69(4):679–699
Chiou JM, Li PL (2008) Correlation-based functional clustering via subspace projection. J Am Stat Assoc 103(484):1684–1692
Coffey N, Hinde J, Holian E (2014) Clustering longitudinal profiles using P-splines and mixed effects models applied to time-course gene expression data. Comput Stat Data Anal 71:14–29
Daawin P, Kim S, Miljkovic T (2019) Predictive modeling of obesity prevalence for the us population. N Am Actuar J 23(1):64–81
de Amorim RC (2012) Constrained clustering with minkowski weighted k-means. In: 2012 IEEE 13th international symposium on computational intelligence and informatics (CINTI). IEEE, pp. 13–17
De Boor C (2001) A practical guide to splines. Springer, New York, NY
Fan J, Li R (2001) Variable selection via nonconcave penalized likelihood and its oracle properties. J Am Stat Assoc 96(456):1348–1360
Fang K, Chen Y, Ma S, Zhang Q (2022) Biclustering analysis of functionals via penalized fusion. J Multivar Anal 189:104874
Foulds J, Kumar S, Getoor L (2015) Latent topic networks: a versatile probabilistic programming framework for topic models. In International conference on machine learning. PMLR, pp. 777–786
Hales CM, Carroll MD, Fryar CD, Ogden CL (2017) Prevalence of obesity among adults and youth: United states, 2015–2016. NCHS data brief (288)
Huang H, Li Y, Guan Y (2014) Joint modeling and clustering paired generalized longitudinal trajectories with application to cocaine abuse treatment data. J Am Stat Assoc 109(508):1412–1424
Hubert L, Arabie P (1985) Comparing partitions. J Classif 2(1):193–218
Ibrahim JG, Zhu H, Tang N (2008) Model selection criteria for missing-data problems using the EM algorithm. J Am Stat Assoc 103(484):1648–1658
Jacques J, Preda C (2013) Funclust: a curves clustering method using functional random variables density approximation. Neurocomputing 112:164–171
Jacques J, Preda C (2014) Functional data clustering: a survey. Adv Data Anal Classif 8(3):231–255
Jain AK (2010) Data clustering: 50 years beyond K-means. Pattern Recogn Lett 31(8):651–666
James GM, Hastie TJ, Sugar CA (2000) Principal component models for sparse functional data. Biometrika 87(3):587–602
James GM, Sugar CA (2003) Clustering for sparsely sampled functional data. J Am Stat Assoc 98(462):397–408
Jiang H, Serban N (2012) Clustering random curves under spatial interdependence with application to service accessibility. Technometrics 54(2):108–119
Li T, Song X, Zhang Y, Zhu H, Zhu Z (2021) Clusterwise functional linear regression models. Comput Stat Data Anal 158:107192
Li Y, Hsing T (2010) Uniform convergence rates for nonparametric regression and principal component analysis in functional/longitudinal data. Ann Stat 38(6):3321–3351
Li Y, Wang N, Carroll RJ (2013) Selecting the number of principal components in functional data. J Am Stat Assoc 108(504):1284–1294
Luan Y, Li H (2003) Clustering of time-course gene expression data using a mixed-effects model with B-splines. Bioinformatics 19(4):474–482
Lv Y, Zhu X, Zhu Z, Qu A (2020) Nonparametric cluster analysis on multiple outcomes of longitudinal data. Stat Sin 30(4):1829–1856
Ma H, Liu C, Xu S, Yang J (2023) Subgroup analysis for functional partial linear regression model. Can J Stat 51(2):559–579
Ma S, Huang J (2017) A concave pairwise fusion approach to subgroup analysis. J Am Stat Assoc 112(517):410–423
Ma S, Huang J, Zhang Z, Liu M (2020) Exploration of heterogeneous treatment effects via concave fusion. Int J Biostat 16(1):20180026. https://www.degruyter.com/document/doi/10.1515/ijb-2018-0026/html
Miljkovic T, Wang X (2021) Identifying subgroups of age and cohort effects in obesity prevalence. Biom J 63(1):168–186
Ng SK, McLachlan GJ, Wang K, Ben-Tovim Jones L, Ng SW (2006) A mixture model with random-effects components for clustering correlated gene-expression profiles. Bioinformatics 22(14):1745–1752
Peng J, Müller HG (2008) Distance-based clustering of sparsely observed stochastic processes, with applications to online auctions. Ann Appl Stat 2(3):1056–1077
Ramsay JO, Silverman BW (2005) Functional data analysis. Springer, New York
Rand WM (1971) Objective criteria for the evaluation of clustering methods. J Am Stat Assoc 66(336):846–850
Redd A (2012) A comment on the orthogonalization of B-spline basis functions and their derivatives. Stat Comput 22(1):251–257
Ren M, Zhang S, Zhang Q, Ma S (2022) Gaussian graphical model-based heterogeneity analysis via penalized fusion. Biometrics 78(2):524–535
Sangalli LM, Secchi P, Vantini S, Vitelli V (2010) K-mean alignment for curve clustering. Comput Stat Data Anal 54(5):1219–1233
Sugar CA, James GM (2003) Finding the number of clusters in a dataset: an information-theoretic approach. J Am Stat Assoc 98(463):750–763
Tibshirani R (1996) Regression shrinkage and selection via the lasso. J R Stat Soc Ser B (Methodol) 58(1):267–288
Vinh NX, Epps J, Bailey J (2010) Information theoretic measures for clusterings comparison: variants, properties, normalization and correction for chance. J Mach Learn Res 11:2837–2854
Wang H, Li R, Tsai CL (2007) Tuning parameter selectors for the smoothly clipped absolute deviation method. Biometrika 94(3):553–568
Wang X, Zhu Z, Zhang HH (2023) Spatial heterogeneity automatic detection and estimation. Comput Stat Data Anal 180:107667
Xiao P, Wang G (2022) Partial functional linear regression with autoregressive errors. Commun Stat Theory Methods 51(13):4515–4536
Yao F, Müller HG, Wang JL (2005) Functional data analysis for sparse longitudinal data. J Am Stat Assoc 100(470):577–590
Zhang CH (2010) Nearly unbiased variable selection under minimax concave penalty. Ann Stat 38(2):894–942
Zhang X, Zhang Q, Ma S, Fang K (2022) Subgroup analysis for high-dimensional functional regression. J Multivar Anal 192:105100
Zhou L, Huang JZ, Carroll RJ (2008) Joint modelling of paired sparse functional data using principal components. Biometrika 95(3):601–619
Zhou L, Sun S, Fu H, Song PXK (2022) Subgroup-effects models for the analysis of personal treatment effects. Ann Appl Stat 16(1):80–103
Zhu X, Qu A (2018) Cluster analysis of longitudinal profiles with subgroups. Electron J Stat 12(1):171–193
Zhu X, Tang X, Qu A (2021) Longitudinal clustering for heterogeneous binary data. Stat Sin 31(2):603–624
Zhu Y, Di C, Chen YQ (2019) Clustering functional data with application to electronic medication adherence monitoring in HIV prevention trials. Stat Biosci 11(2):238–261
Funding
This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.
Author information
Authors and Affiliations
Corresponding author
Ethics declarations
Conflict of interest
No financial or non-financial interests are directly or indirectly related to the submitted work.
Additional information
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Appendix
Appendix
In this appendix, the EM algorithm with a known group structure is presented. The EM procedure is similar to the EM algorithm in James et al. (2000), the main difference is that a new design matrix is constructed based on the given group information.
If the group structure is known, suppose there are \(\tilde{K}\) groups and define \(\tilde{\varvec{W}}\) be an \(n\times \tilde{K}\) matrix with element \(w_{ij}\) and \(w_{ij}=1\) if i is in the kth group. Also define \(\varvec{W}=\tilde{\varvec{W}}\otimes \varvec{I}_{q}\) and \(\varvec{U}=\varvec{B}_{0}\varvec{W}\). \(\left( \tilde{\varvec{\alpha }}_{1}^{T},\dots ,\tilde{\varvec{\alpha }}_{\tilde{K}}^{T}\right) ^{T}=\tilde{\varvec{\alpha }}=\left( \varvec{U}^{T}\varvec{U}\right) ^{-1}\varvec{U}^{T}\varvec{Y}\) is the estimate of coefficients for \(\tilde{K}\) groups \(\varvec{\alpha } = (\varvec{\alpha }_1^T,\dots , \varvec{\alpha }_{\tilde{K}}^T)^T\), which is set as the initial estimate of \(\varvec{\alpha }\). Thus, \(\tilde{\varvec{\beta }}_{i}=\tilde{\varvec{\alpha }}_{k}\) if i is in the kth group. Define
where \(\varvec{\beta }_i^*\) is obtained using the same procedure in Remark 2. Then, the eigendecomposition is done for \(\varvec{C}_{n}=\varvec{\Theta }_{0}\varvec{\Lambda }_{0}\varvec{\Theta }_{0}^{T}\), where \(\varvec{\Theta }_0\) and \(\varvec{\Lambda }_0\) are the initial values of \(\varvec{\Theta }\) and \(\varvec{\Lambda }\), respectively.
Similar to the proposed algorithm, the conditional distribution of \(\varvec{\xi }_i\) is needed, which has the following forms
where \(\varvec{m}_{i} = \; { E\left[ \varvec{\xi }_{i}\vert \varvec{\alpha }, \varvec{\Theta }, \varvec{\lambda }, \sigma ^2 \right] }\) and \(\varvec{V}_{i} = \; { V\left[ \varvec{\xi }_{i}\vert \varvec{\alpha }, \varvec{\Theta }, \varvec{\lambda }, \sigma ^2 \right] }\) with the following form.
The only difference between the conditional distribution here and the proposed algorithm is that \(\varvec{\alpha }\) is used instead of \(\varvec{\beta }\), since the group structure information is given.
Similarly, \(\sigma ^2\) is updated by
Also, the same procedure is used to updated \(\varvec{\Theta }\) and \(\varvec{\lambda }\) with
Last, \(\varvec{\alpha }\) is updated as
Rights and permissions
Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.
About this article
Cite this article
Wang, X. Clustering of longitudinal curves via a penalized method and EM algorithm. Comput Stat 39, 1485–1512 (2024). https://doi.org/10.1007/s00180-023-01380-2
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s00180-023-01380-2