Classifying patients according to longitudinal measures, or trajectory classification, has become frequent in clinical research. The k-means algorithm is increasingly used for this task in case of continuous variables with standard deviations that do not depend on the mean. One feature of count and binary data modeled by Poisson or logistic regression is that the variance depends on the mean; hence, the within-group variability changes from one group to another depending on the mean trajectory level. Mixture modeling could be used here for classification though its main purpose is to model the data. The results obtained may change according to the main objective. This article presents an extension of the k-means algorithm that takes into account the features of count and binary data by using the deviance as distance metric. This approach is justified by its analogy with the classification likelihood. Two applications are presented with binary and count data to show the differences between the classifications obtained with the usual Euclidean distance versus the deviance distance.
Proust-LimaCSéneMTaylorJM. Joint latent class models for longitudinal and time-to-event data: a review. Stat Methods Med Res2014; 23(1): 74–90.
2.
BastardMSoulinphumyKPhimmasoneP. Women experience a better long-term immune recovery and a better survival on HAART in Lao People's Democratic Republic. BMC Infect Dis2013; 22(13): 27–27.
MacQueen JB. Some methods for classification and analysis of multivariate observations. In: Proceedings of 5-th Berkeley symposium on mathematical statistics and probability, Berkeley, University of California Press, 1967.
5.
JacquesJPredaC. Functional data clustering: a survey, Lille: INRIA, 2013.
6.
GenoliniCFalissardB. KmL: a package to cluster longitudinal data. Comput Methods Prog Biomed2011; 104: e112–e121.
7.
AltHGodauM. Computing the Fréchet distance between two polygonal curves. Int J Comput Geom Appl1995; 5: 75–91.
8.
WahlSKrugSThenC. Comparative analysis of plasma metabolomics response to metabolic challenge tests in healthy subjects and influence of the FTO obesity risk allele. Metabolomics2014; 10(3): 386–401.
9.
RancièreFNikasinovicLBousquetJ. Onset and persistence of respiratory/allergic symptoms in preschoolers: new insights from the PARIS birth cohort. Allergy2013; 68(9): 1158–1167.
10.
PingaultJBTremblayREVitaroF. Childhood trajectories of inattention and hyperactivity and prediction of educational attainment in early adulthood: a 16-year longitudinal population-based study. Am J Psychiatr2011; 168: 1164–1170.
11.
BrossartDFParkerRWillsonVL. A comparison of two methods for analyzing longitudinal data: tuckerized growth curves and an application of K means analysis. Learn Individ Differ1998; 10(2): 121–136.
12.
JamesGMSugarCA. Clustering for sparsely sampled functional data. J Am Stat Assoc2003; 98(462): 397–408.
13.
BouveyronCGirardSSchmidC. High dimensional data clustering. Comput Stat Data Anal2007; 52: 502–519.
14.
JacquesJPredaC. Model-based clustering for multivariate functional data. Comput Stat Data Anal2014; 71: 92–106.
15.
ChuM-KMKovalJJ. Trajectory modelling of longitudinal binary data: application of the EM algorithm for mixture models. Commun Stat Simul Comput2014; 43(3): 495–519.
16.
NaginDS. Group-based modeling of development, London: Harvard University Press, 2005.
17.
PicklesACroudaceT. Latent mixture models for multivariate and longitudinal outcomes. Stat Methods Med Res2009; 19(3): 271–289.
18.
VrbikIMcNicholasPD. Parsimonious skew mixture models for model-based clustering and classification. Comput Stat Data Anal2014; 71: 196–210.
19.
JonesBLNaginDS. Advances in group-based trajectory modeling and an SAS procedure for estimating them. Sociol Methods Res2007; 35: 542–571.
20.
MuthénBOSheddenK. Finite mixture modeling with mixture outcomes using the EM algorithm. Biometrics1999; 55(2): 463–469.
21.
CeleuxGGovaertG. Comparison of the mixture and the classification maximum likelihood. J Stat Comput Simul1993; 47: 127–146.
22.
HathawayRJ. Another interpretation of the EM algorithm for mixture distributions. Stat Probab Lett1986; 4: 53–56.
23.
HartiganJAWongMA. Algorithm AS 136: a K-means clustering algorithm. J R Stat Soc Ser C Appl Stat1979; 28(1): 100–108.
24.
KanungoTMountDMNetanyahuNS. An efficient k-means clustering algorithm: analysis and implementation. IEEE Trans Pattern Anal Mach Intell2002; 24(7): 881–892.
25.
SymonsMJ. Clustering criteria and multivariate normal mixtures. Biometrics1981; 37: 35–43.
26.
CeleuxGGovaertG. A classification EM algorithm for clustering and two stochastic versions. Comput Stat Data Anal1992; 14: 315–332.
27.
ElsensohnMHKlichAEcochardR. A graphical method to assess distribution assumption in group-based trajectory models. Stat Methods Med ResEpub ahead of print 2013.
28.
CeleuxGGovaertG. Clustering criteria for discrete data and latent class models. J Classif1991; 8: 157–176.
29.
McCullaghPNelderJA. Generalized linear models, 2nd ed. Boca Raton: Chapman and Hall/CRC, 1989.
30.
CorbelVAkogbetoMDamienGB. Combination of malaria vector control interventions in pyrethroid resistance area in Benin: a cluster randomised controlled trial. Lancet Infect Dis2012; 12(8): 617–626.
31.
DesclauxALanièceINdoyeI. The Senegalese antiretroviral drug access initiative. An economic, social, behavioural and biomedical analysis, Paris: ANRS, UNAIDS, WHO, 2004.
32.
BastardMFallMBLanièceI. Revisiting long-term adherence to highly active antiretroviral therapy in Senegal using latent class analysis. J Acquir Immune Defic Syndr2011; 57(1): 55–61.
33.
GenoliniCEcochardRJacqmin-GaddaH. Copy mean: a new method to impute intermittent missing values in longitudinal studies. Open J Stat2013; 3: 26–40.
34.
GovaertGNadifM. Comparison of the mixture and the classification maximum likelihood in cluster analysis with binary data. Comput Stat Data Anal1996; 23: 65–81.
35.
BiernackiCGovaertG. Using the classification likelihood to choose the number of clusters. Comput Sci Stat1997; 29(2): 451–457.
36.
McLachlanGPeelD. Finite mixture models, New York, NY: Wiley, 2000.
37.
BiernackiCCeleuxGGovaertG. Assessing a mixture model for clustering with the integrated classification likelihood, Grenoble: IS2 (INRIA Rhône Alpes), 1998.
38.
BiernackiCGovaertG. Choosing models in model-based clustering and discriminant analysis. J Stat Comput Simul1999; 64: 49–71.
39.
NaginDSOdgersCL. Group-based trajectory modeling in clinical research. Annu Rev Clin Psychol2010; 6: 109–138.
40.
NylundKLAsparouhovTMuthénBO. Deciding on the number of classes in latent class analysis and growth mixture modeling: a Monte Carlo simulation study. Struct Equ Modeling2007; 4(4): 535–569.
41.
Muthén LK Muthén BO Mplus User's Guide. 7th ed. Los Angeles: CA Muthén & Muthén, 2012.
42.
ProustCJacqmin-GaddaH. Estimation of linear mixed models with a mixture of distribution for the random effects. Comput Methods Prog Biomed2005; 78(2): 165–173.
43.
Proust-LimaCAmievaHJacqmin-GaddaH. Analysis of multivariate mixed longitudinal data: a flexible latent process approach. Br J Math Stat Psychol2013; 66(3): 470–487.
44.
BryantPWilliamsonJA. Asymptotic behaviour of classification maximum likelihood estimates. Biometrika1978; 65(2): 273–281.
