Mixed Hidden Markov Model for Heterogeneous Longitudinal Data with Missingness and Errors in the Outcome Variable Titre: Modèle de Markov caché mixte pour des données longitudinales hétérogènes avec erreurs et données manquantes dans la variable de sortie

作者: Cyrille Delpierre , Nicolas Savy , Thierry Lang , Sébastien Gadat , Dominique Dedieu

DOI:

关键词:

摘要: Analysing longitudinal declarative data raises many difficulties, such as the processing of errors and missingness in outcome variable. Moreover, long-term monitored cohorts (commonly encountered life-course epidemiology) may reveal a problem time heterogeneity, especially regarding way subjects respond to investigator. We propose Mixed Hidden Markov Model which considers several causes randomness response also enables effect past health act on present responses through memory state. Hence, we take into account both missing responses, retrospective questions. thus Stochastic Expectation Maximization algorithm (SEM), is less time-consuming than usual EM algorithms perform estimation parameters our MHMM. carry out simulation study assess performances this context cancer epidemiology with British NCDS 1958 cohort. Simulations show that covariates transitions probabilities estimated moderate bias. At last, investigate brief real application early social class smoking behaviour. It appears female sample used, does not mainly behaviours. more information needed compensate for view improve statistical analysis. Resume : L'analyse de donnees declaratives longitudinales fait apparaitre nombreuses difficultes, comme le traitement des erreurs et manquantes la variable sortie. En outre, les cohortes suivies sur long terme, telles que celles utilisees en epidemiologie "life-course" peuvent soulever un probleme d'heterogeneite du temps, surtout ce qui concerne facon repondre aux questions l'enqueteur. Nous proposons dans cet article l'introduction d'un modele cache mixte comprend possibilites d'erreur non-reponse, permet egalement considerer l'effet resultat sante passe peut agir reponses actuelles travers une memoire d' etat. estimations, nous avons d'utiliser algorithme Stochastique est moins gourmand temps calcul l'algorithme usuel utilisant integration effets aleatoires. effectue etude par afin d'evaluer contexte l'epidemiologie avec cohorte britanniques "NCDS 1958". Les simulations ont montre covariables probabilites ete estimee biais modere. Enfin, realise reelles etudiant classe sociale precoce comportement tabagique. Il apparu que, l'echantillon femmes utilise pour cette enquete, n'agit pas principalement l'usage tabac. Cependant, plus d'information necessaire compenser declaration obtenir meilleurs resultats statistiques.

参考文章(36)
D. Commenges, Multi-state Models in Epidemiology Lifetime Data Analysis. ,vol. 5, pp. 315- 327 ,(1999) , 10.1023/A:1009636125294
Hakon K Gjessing, Odd O Aalen, Oernulf Borgan, Survival and Event History Analysis: A Process Point of View ,(2008)
Eric Moulines, Marc Lavielle, Bernard Delyon, Convergence of a stochastic approximation version of the EM algorithm Annals of Statistics. ,vol. 27, pp. 94- 128 ,(1999) , 10.1214/AOS/1018031103
Applied Latent Class Analysis Canadian Journal of Sociology-cahiers Canadiens De Sociologie. ,vol. 28, pp. 584- ,(2002) , 10.1017/CBO9780511499531
Maud Delattre, Inference in Mixed Hidden Markov Models and Applications to Medical Studies Journal de la Société Française de Statistique & revue de statistique appliquée. ,vol. 151, pp. 90- 105 ,(2010)
Alexandre Bureau, Stephen Shiboski, James P. Hughes, Applications of continuous time hidden Markov models to the study of misclassified disease outcomes Statistics in Medicine. ,vol. 22, pp. 441- 462 ,(2003) , 10.1002/SIM.1270
M. Delattre, M. Lavielle, Maximum likelihood estimation in discrete mixed hidden Markov models using the SAEM algorithm Computational Statistics & Data Analysis. ,vol. 56, pp. 2073- 2085 ,(2012) , 10.1016/J.CSDA.2011.12.017
Robert J Tibshirani, Bradley Efron, An introduction to the bootstrap ,(1993)
Rachel MacKay Altman, Mixed Hidden Markov Models Journal of the American Statistical Association. ,vol. 102, pp. 201- 210 ,(2007) , 10.1198/016214506000001086
Michelle Kelly-Irving, Benoit Lepage, Dominique Dedieu, Rebecca Lacey, Noriko Cable, Melanie Bartley, David Blane, Pascale Grosclaude, Thierry Lang, Cyrille Delpierre, Childhood adversity as a risk for cancer: findings from the 1958 British birth cohort study BMC Public Health. ,vol. 13, pp. 767- 767 ,(2013) , 10.1186/1471-2458-13-767