TY - JOUR
T1 - Bootstrap validation of the estimated parameters in mixture models used for clustering
JF - Journal de la Société Française de Statistique
Y1 - 2019
A1 - Taushanov, Zhivko
A1 - Berchtold, André
KW - www
AB - When a mixture model is used to perform clustering, the uncertainty is related both to the choice of an optimal model (including the number of clusters) and to the estimation of the parameters. We discuss here the computation of confidence intervals using different bootstrap approaches, which either mix or separate the two kinds of uncertainty. In particular, we suggest two new approaches that rely to some degree on the model specification considered as optimal by the researcher, and that address specifically the uncertainty related to parameter estimation. These methods are especially useful for poorly separated data or complex models, where the selected solution is difficult to recreate in each bootstrap sample, and they present the advantage of reducing the well-known label-switching issue. Two simulation experiments based on the Hidden Mixture Transition Distribution model for the clustering of longitudinal data illustrate our proposed bootstrap approaches.
VL - 160
UR - http://journal-sfds.fr/article/view/730
ER -
TY - JOUR
T1 - Does the primary resource of sex education matter? A Swiss national study
JF - The Journal of Sex Research
Y1 - 2019
A1 - Barrense-Dias, Yara
A1 - Akre, Christina
A1 - Suris, Joan-Carles
A1 - Berchtold, André
A1 - Morselli, Davide
A1 - Jacot-Descombes, Caroline
A1 - Leeners, Brigitte
KW - www2
Y1 - jun
UR - https://www.tandfonline.com/doi/full/10.1080/00224499.2019.1626331
PY - 10.1080/00224499.2019.1626331
ER -
TY - JOUR
T1 - Illustrating instrumental variable regressions using the career adaptability – job satisfaction relationship
JF - Frontiers in Psychology
Y1 - 2019
A1 - Bollmann, Grégoire
A1 - Rouzinov, Serguei
A1 - Berchtold, André
A1 - Rossier, Jérôme
KW - Affect (emotion
KW - career adaptability
KW - causal inference
KW - instrumental variable (IV)
KW - job satisfaction
KW - mood
KW - personality
KW - personality)
KW - www
AB - This article illustrates instrumental variable (IV) estimation by examining an unexpected finding of the research on career adaptability and job satisfaction. Theoretical and empirical arguments suggest that in the general population, people’s abilities to adapt their careers are beneficial to their job satisfaction. However, a recent meta-analysis unexpectedly found no effect when personality traits are controlled for. We argue that a reverse effect of job satisfaction on career adaptability, originating from affective tendencies tied to personality, might explain this null effect. Our argument implies that the estimates obtained with traditional ordinary least squares (OLS) regressions are biased by endogeneity, a correlation between an explanatory variable and the error term in a regression model. When experimental manipulations are impossible, IV estimations, such as two-stage least squares (2SLS) regressions, are one possible solution to the endogeneity problem. Analyzing three waves of data from a sample of 836 adults, the concurrent and time-lagged effect of job satisfaction on career adaptability was revealed to be more consistent than the reverse. Our results provide an explanation, rooted in affective dispositions, as to why recent meta-analytical estimates unexpectedly found that career adaptability does not predict job satisfaction at the interindividual level. We also discuss IV estimation in terms of its limits, weight the interpretation of its estimates against the temporality criterion for causal inference, and consider its possible extension to analyses of change.
VL - 10
UR - https://www.frontiersin.org/articles/10.3389/fpsyg.2019.01481/full
PY - 10.3389/fpsyg.2019.01481
ER -
TY - JOUR
T1 - Daily internet time: towards an evidence-based recommendation?
JF - European Journal of Public Health
Y1 - 2018
A1 - Berchtold, André
A1 - Akre, Christina
A1 - Barrense-Dias, Yara
A1 - Zimmermann, Grégoire
A1 - Suris, Joan-Carles
KW - adolescent
KW - child
KW - evidence-based practice
KW - health outcomes
KW - internet
KW - screen time
AB - Background: Since 2001, a recommendation of no more than 2 h per day of screen time for children 2 years of age or older was adopted in many countries. However, this recommendation was rarely examined empirically. The goal of the present study was to question this recommendation in today’s connected world. Methods: We used data from the ado@internet.ch survey (spring 2012), a representative sample of 8th graders in the Canton of Vaud, Switzerland (n = 2942, 50.6% female). Internet use, health outcomes, substance use, well-being and socio-demographic characteristics were considered. Bi-variate statistical analyses were performed. Results: All outcomes were significantly associated with the time spent on internet, more time being associated with a higher prevalence of adverse consequences. Youth spending on average one more hour on Internet per day than the reference category (1.5–2.5 h) did not differ in terms of adverse health outcomes. Differences began to appear on sleeping problems, tobacco use, alcohol misuse, cannabis use and sport inactivity with youth spending between 3.5 h and 4.5 h per day on internet. Conclusions: This study demonstrates the absence of justification for setting a limit to only 2 h of screen time per day. Significant effects on health seem to appear only beyond 4 h per day and there may be benefits for those who spend less than an hour and a half on internet.
UR - https://academic.oup.com/eurpub/advance-article/doi/10.1093/eurpub/cky054/4973864
J1 - Daily internet time
PY - 10.1093/eurpub/cky054
ER -
TY - JOUR
T1 - The heterogeneity of disability trajectories in later life: Dynamics of activities of daily living performance among nursing home residents
JF - Journal of Aging & Health
Y1 - 2018
A1 - Bolano, Danilo
A1 - Berchtold, André
A1 - Burge, Elisabeth
KW - ADL trajectories
KW - longitudinal analysis
KW - multi-state model
KW - nursing home resident
KW - variability disability trajectories
AB - Objective: This study investigated the variability in activities of daily living (ADL) trajectories among 6,155 nursing home residents using unique and rich observational data. Method: The impairment in ADL performance was considered as a dynamic process in a multi-state framework. Using an innovative mixture model, such states were not defined a priori but inferred from the data. Results: The process of change in functional health differed among residents. We identified four latent regimes: stability or slight deterioration, relevant change, variability, and recovery. Impaired body functions and poor physical performance were main risk factors associated with degradation in functional health. Discussion: The evolution of disability in later life is not completely gradual or homogeneous. Steep deterioration in functional health can be followed by periods of stability or even recovery. The current condition can be used to successfully predict the evolution of ADL allowing to set and target different care priorities and practices.
UR - http://journals.sagepub.com/doi/abs/10.1177/0898264318776071
J1 - The heterogeneity of disability trajectories in later life
PY - 10.1177/0898264318776071
ER -
TY - CHAP
T1 - Markovian-based clustering of internet addiction trajectories
T2 - Sequence analysis and related approaches : Innovative methods and applications.
Y1 - 2018
A1 - Taushanov, Zhivko
A1 - Berchtold, André
ED - Ritschard, Gilbert
ED - Studer, Matthias
JA - Sequence analysis and related approaches : Innovative methods and applications.
PB - Springer
CY - Cham, Switzerland
UR - https://link.springer.com/chapter/10.1007/978-3-319-95420-2_12
PY - 10.1007/978-3-319-95420-2_12
ER -
TY - JOUR
T1 - A direct local search method and its application to a markovian model
JF - Statistics, Optimization & Information Computing
Y1 - 2017
A1 - Taushanov, Zhivko
A1 - Berchtold, André
KW - heuristic
KW - Hidden Mixture Transition Distribution (HMTD) model
KW - hill-climbing method
KW - longitudinal data
KW - optimization
AB - While the hidden mixture transition distribution (HMTD) model is a powerful framework for the description, analysis, and classification of longitudinal sequences of continuous data, it is notoriously difficult to estimate because of the complexity of its solution space. In this paper, we explore how a new heuristic specifically developed for the HMTD performs compared to different standard optimization algorithms. This specific heuristic can be classified as a hill-climbing method, and different variants are proposed, including a jittering procedure to escape local maxima and measures to speed up the convergence.Different popular approaches are used for comparison, including PSO, SA, GA, NM, L-BFGS-B, and DE. The same HMTD model was optimized on different datasets and the results were compared in terms of both fit to the data and estimated parameters. Even if the complexity of the problem implies that no one algorithm can be considered as an overall best, our heuristic performed well in all situations, leading to useful solutions in terms of both fit and interpretability.
VL - 5
Y1 - 03/2017
UR - http://www.iapress.org/index.php/soic/article/view/20170302
CP - 1
PY - 10.19139/soic.v5i1.253
ER -
TY - JOUR
T1 - Imputation of repeatedly observed multinomial variables in longitudinal surveys
JF - Communications in Statistics - Simulation and Computation
Y1 - 2017
A1 - Berchtold, André
A1 - Suris, Joan-Carles
KW - 62P25
KW - causality
KW - chained equations
KW - longitudinal survey
KW - missing data
KW - multiple imputation
KW - Primary 62
KW - Secondary 62D05
AB - It is now a standard practice to replace missing data in longitudinal surveys with imputed values, but there is still much uncertainty about the best approach to adopt. Using data from a real survey, we compared different strategies combining multiple imputation and the chained equations method, the two main objectives being (1) to explore the impact of the explanatory variables in the chained regression equations and (2) to study the effect of imputation on causality between successive waves of the survey. Results were very stable from one simulation to another, and no systematic bias did appear. The critical points of the method lied in the proper choice of covariates and in the respect of the temporal relation between variables.
VL - 46
UR - https://www.tandfonline.com/doi/full/10.1080/03610918.2015.1082588
CP - 4
PY - 10.1080/03610918.2015.1082588
ER -
TY - CONF
T1 - Comparison of two bootstrap procedures in the case of hidden Markovian model clustering
Y1 - 2016
A1 - Taushanov, Zhivko
A1 - Berchtold, André
KW - bootstrap
KW - clustering
KW - hidden Markov model
KW - mixture models
AB - The clustering of longitudinal data sequences is considered using a latent Markovian model (HMTD) combining Gaussian distributions and covariates. The main objective is to evaluate the significance of the estimated parameters. At first, different model specifications are optimized and the one providing the best clustering in terms of BIC is selected. Two different bootstrap procedures are then applied and compared in order to investigate the significance of the parameters of this optimal solution. First, a standard bootstrap procedure is applied using the full original sample and the optimal model with multiple components (clusters) is computed at each iteration. That leads to solutions with different degrees of similarity with the optimal solution and the well-known label-switching problem may occur. An alternative procedure is proposed that consists in applying separate bootstrap procedures on each subsample defined by the optimal clustering. In this case, a single component model is estimated from each bootstrap iteration and for each cluster separately. This method also provides a confidence interval for each parameter and avoids the label-switching problem. The pros and cons of each approach are described and examples based on real data are provided.
PB - European Regional Section of the IASC
CY - Oviedo, Spain
UR - http://www.compstat2016.org/
ER -
TY - CHAP
T1 - A discussion on hidden Markov models for life course data
T2 - Proceedings of the International Conference on Sequence Analysis and Related Methods (LaCOSA II)
Y1 - 2016
A1 - Bolano, Danilo
A1 - Berchtold, André
A1 - Ritschard, Gilbert
KW - hidden Markov model
KW - Life course approach
KW - sequence analysis
AB - This is an introduction on discrete-time Hidden Markov models (HMM) for longitudinal data analysis in population and life course studies. In the Markovian perspective, life trajectories are considered as the result of a stochastic process in which the probability of occurrence of a particular state or event depends on the sequence of states observed so far. Markovian models are used to analyze the transition process between successive states. Starting from the traditional formulation of a first-order discrete-time Markov chain where each state is liked to the next one, we present the hidden Markov models where the current response is driven by a latent variable that follows a Markov process. The paper presents also a simple way of handling categorical covariates to capture the effect of external factors on the transition probabilities and existing software are briefly overviewed. Empirical illustrations using data on self reported health demonstrate the relevance of the different extensions for life course analysis.
JA - Proceedings of the International Conference on Sequence Analysis and Related Methods (LaCOSA II)
PB - NCCR LIVES
CY - Lausanne, Switzerland
UR - https://lacosa.lives-nccr.ch/sites/lacosa.lives-nccr.ch/files/lacosa2-proceedings.pdf
ER -
TY - JOUR
T1 - General framework and model building in the class of Hidden Mixture Transition Distribution models
JF - Computational Statistics & Data Analysis
Y1 - 2016
A1 - Bolano, Danilo
A1 - Berchtold, André
KW - BIC
KW - hidden Markov model
KW - mixture model
KW - mixture transition distribution model
KW - model selection
KW - panel data
AB - Modeling time series that present non-Gaussian features plays as central role in many fields, including finance, seismology, psychological, and life course studies. The Hidden Mixture Transition Distribution model is an answer to the complexity of such series. The observed heterogeneity can be induced by one or several latent factors, and each level of these factors is related to a different component of the observed process. The time series is then treated as a mixture and the relation between the components is governed by a Markovian latent transition process. This framework generalizes several specifications that appear separately in related literature. Both the expectation and the standard deviation of each component are allowed to be functions of the past of the process. The latent process can be of any order, and can be modeled using a discrete Mixture Transition Distribution. The effects of covariates at the visible and hidden levels are also investigated. One of the main difficulties lies in correctly specifying the structure of the model. Therefore, we propose a hierarchical model selection procedure that exploits the multilevel structure of our approach. Finally, we illustrate the model and the model selection procedure through a real application in social science.
VL - 93
Y1 - 01/2016
PY - 10.1016/j.csda.2014.09.011
ER -
TY - JOUR
T1 - On-line life history calendar and sensitive topics: A pilot study
JF - Computers in Human Behavior
Y1 - 2016
A1 - Morselli, Davide
A1 - Berchtold, Annick
A1 - Joan-Carles Suris Granell
A1 - Berchtold, André
KW - event history calendar
KW - life history calendar
KW - on-line survey
KW - retrospective data collection
KW - sensitive data
AB - The use of the life history calendar (LHC) or the event history calendar as tools for collecting retrospective data has received increasing attention in many fields of social science and medicine. However, little research has examined the use of this method with web-based surveys. In this study, we adapted this method to an on-line setting to collect information about young adults' life histories, sexual behaviors, and substance use. We hypothesized that the LHC method would help respondents to date sensitive and non-sensitive events more precisely than when using a conventional questionnaire. We conducted an experimental design study comparing university students' responses to an on-line LHC and a conventional on-line question list. A test-retest design in which the respondents completed the survey again two weeks later was also applied to test the precision and reliability of the participants' dating of events. The results showed that whereas the numbers of sensitive and non-sensitive events were generally similar for the two on-line questionnaires, the responses obtained with the LHC were more consistent across the two administrations. Analyses of the respondents' on-line behavior while completing the LHC confirmed that respondents used the LHC's graphic interface to correct and reedit previous answers, thus decreasing data errors.
VL - 58
Y1 - 05/2016
PY - 10.1016/j.chb.2015.12.068
ER -
TY - JOUR
T1 - Test–retest: Agreement or reliability?
JF - Methodological Innovations
Y1 - 2016
A1 - Berchtold, André
KW - agreement
KW - Bland–Altman plot
KW - concordance coefficient
KW - correlation
KW - reliability
KW - Test–retest
AB - Test–retest is a concept that is routinely evaluated during the validation phase of many measurement tools. However, this term covers at least two related but very different concepts: reliability and agreement. Reliability is the ability of a measure applied twice upon the same respondents to produce the same ranking on both occasions. Agreement requires the measurement tool to produce twice the same exact values. An analysis of research papers showed that the distinction between both concepts remains anything but clear, and that the current practice is to evaluate reliability only, generally on the basis of the sole Pearson’s correlation. This practice is very problematic in the context of longitudinal studies because it becomes difficult to determine whether a difference between two successive observations is attributable to a real change of the respondents or only to the characteristics of the measurement tool, which then leads to a possible misinterpretation of the results. More focus should be given on the real interpretation of linear correlation, and when agreement is required in addition to reliability, then correct alternative, such as the Bland–Altman plot, should be more generally used.
VL - 9
Y1 - oct
PY - 10.1177/2059799116672875
ER -
TY - CHAP
T1 - Using dynamic microsimulation to understand professional trajectories of the active Swiss population
T2 - Proceedings of the International Conference on Sequence Analysis and Related Methods (LaCOSA II)
Y1 - 2016
A1 - Adamopoulos, Pauline
A1 - Ritschard, Gilbert
A1 - Berchtold, André
KW - longitudinal data
KW - Markov models
KW - microsimulation
KW - professional trajectories
AB - Within the social and economic sciences and of particular interest to demographers are life course events. Looking at life sequences we can better understand which states, or life events, precede or are precursors to vulnerability. A tool that has been used for policy evaluation and recently has been gaining ground in life course sequence simulation is dynamic microsimulation. Within this context dynamic microsimulation consists in generating entire life courses from the observation of portions of the trajectories of individuals of different ages. In this work, we aim to use dynamic microsimulation in order to analyse individual professional trajectories with a focus on vulnerability. The primary goal of this analysis is to deepen upon current literature by providing insight from a longitudinal perspective on the signs of work instability and the process of precarity. The secondary goal of this work which is to show how, by using microsimulation, data collected for one purpose can be analysed under a different scope and used in a meaningful way. The data to be used in this analysis are longitudinal and were collected by NCCR-LIVES IP207 under the supervision of Prof. Christian Maggiori and Dr. Gregoire Bollmann. Individuals aged 25 to 55 residing in the German-speaking and French-speaking regions of Switzerland were followed annually for four years. These individuals were questioned regarding, amongst their personal, professional and overall situations and well-being. At the end of the fourth wave, there were 1131 individuals who had participated in all waves. The sample remained representative of the Swiss population with women and the unemployed slightly over represented. Using the information collected from these surveys, we use simulation to construct various longitudinal data modules where each data module represents a specific life domain. We postulate the relationship between these modules and layout a framework of estimation. Within certain data modules a set of equations are created to model the process therein. For every dynamic (time-variant) data module, such as the labour-market module, the transition probabilities between states (ex. labour market status) are estimated using a Markov model and then the possible outcomes are simulated. The benefit of using dynamic microsimulation is that longitudinal sample observations instead of stylised profiles are used to model population dynamics. This is one of the main reasons large-scale dynamic microsimulation models are employed by many developed nations. There has been limited use, however, of such approaches with Swiss data. This work contributes to the analysis of professional trajectories of the active Swiss population by utilising dynamic microsimulation methods.
JA - Proceedings of the International Conference on Sequence Analysis and Related Methods (LaCOSA II)
PB - NCCR LIVES
CY - Lausanne, Switzerland
UR - https://lacosa.lives-nccr.ch/sites/lacosa.lives-nccr.ch/files/lacosa2-proceedings.pdf
ER -
TY - JOUR
T1 - Analyse des événements d’histoire de vie. Estimation de modèles logistiques à temps discret
JF - Cahiers Recherche et Méthodes
Y1 - 2013
A1 - Le Goff, Jean-Marie
A1 - Forney, Yannic
ED - Antonietti, Jean-Philippe
ED - Berchtold, André
PB - University of Lausanne
CY - Lausanne
UR - http://www.unil.ch/webdav/site/consultation-statistique/shared/Cahiers_CREM/CREM_3.pdf
ER -
TY - JOUR
T1 - Méthodes non-paramétriques de l’analyse des événements du parcours de vie (Event history analysis). Estimation avec SPSS. Méthode de Kaplan-Meier et méthode actuarielle
JF - Cahiers Recherche et Méthodes
Y1 - 2013
A1 - Le Goff, Jean-Marie
A1 - Forney, Yannic
ED - Antonietti, Jean-Philippe
ED - Berchtold, André
PB - University of Lausanne
CY - Lausanne
UR - http://www.unil.ch/webdav/site/consultation-statistique/shared/Cahiers_CREM/CREM_2.pdf
ER -