Quantile regression provides a detailed and robust picture of the distribution of a response variable, conditional on a set of observed covariates. Recently, it has be been extended to the analysis of longitudinal continuous outcomes using either time-constant or time-varying random parameters. However, in real-life data, we frequently observe both temporal shocks in the overall trend and individual-specific heterogeneity in model parameters. A benchmark dataset on HIV progression gives a clear example. Here, the evolution of the CD4 log counts exhibits both sudden temporal changes in the overall trend and heterogeneity in the effect of the time since seroconversion on the response dynamics. To accommodate such situations, we propose a quantile regression model, where time-varying and time-constant random coefficients are jointly considered. Since observed data may be incomplete due to early drop-out, we also extend the proposed model in a pattern mixture perspective. We assess the performance of the proposals via a large-scale simulation study and the analysis of the CD4 count data.
GeraciMBottaiM. Quantile regression for longitudinal data using the asymmetric Laplace distribution. Biostatistics2007; 8: 140–154.
2.
LiuYBottaiM. Mixed-effects models for conditional quantiles with longitudinal data. Int J Biostat2009; 5: 1–24.
3.
GeraciMBottaiM. Linear quantile mixed models. Stat Comput2014; 24: 461–479.
4.
TzavidisNSalvatiNSchmidTet al.Longitudinal analysis of the strengths and difficulties questionnaire scores of the Millennium Cohort Study children in England using M-quantile random-effects regression. J R Stat Soc Series A2016; 179: 427–452.
5.
ReichBJBondellHDWangHJ. Flexible Bayesian quantile regression for independent and clustered data. Biostatistics2010; 11: 337–352.
6.
YuanYYinG. Bayesian quantile regression for longitudinal studies with nonignorable missing data. Biometrics2010; 66: 105–114.
7.
BartolucciFFarcomeniA. A multivariate extension of the dynamic logit model for longitudinal data based on a latent Markov heterogeneity structure. J Am Stat Assoc2009; 104: 816–831.
8.
FarcomeniA. Quantile regression for longitudinal data based on latent Markov subject-specific parameters. Stat Comput2012; 22: 141–152.
HardingMLamarcheC. A quantile regression approach for estimating panel data models using instrumental variables. Econ Lett2009; 104: 133–135.
12.
GalvaoAFMontes-RojasGV. Penalized quantile regression for dynamic panel data. J Stat Plan Inference2010; 140: 3476–3497.
13.
GalvaoAF. Quantile regression for dynamic panel data with fixed effects. J Econom2011; 164: 142–157.
14.
LiuMDanielsMJPerriMG. Quantile regression in the presence of monotone missingness with sensitivity analysis. Biostatistics2016; 17: 108–121.
15.
MarinoMFFarcomeniA. Linear quantile regression models for longitudinal experiments: an overview. METRON2015; 73: 229–247.
16.
KaslowRAOstrowDDetelsRet al.The multicenter AIDS cohort study: rationale, organization, and selected characteristics of the participants. Am J Epidemiol1987; 126: 310–318.
17.
ZegerSLDigglePJ. Semiparametric models for longitudinal data with application to CD4 cell numbers in HIV seroconverters. Biometrics1994; 50: 689–699.
18.
FarcomeniAVivianiS. Longitudinal quantile regression in presence of informative drop-out through longitudinal-survival joint modeling. Stat Med2015; 34: 1199–1213.
19.
MarinoMFAlfóM. Latent drop-out based transitions in linear quantile hidden Markov models for longitudinal responses with attrition. Adv Data Anal Classi2015; 9: 483–502.
20.
RoyJ. Modeling longitudinal data with nonignorable dropouts using a latent dropout class model. Biometrics2003; 59: 829–836.
21.
RoyJDanielsM. A general class of pattern mixture models for nonignorable dropout with many possible dropout times. Biometrics2008; 64: 538–545.
22.
YuKZhangJ. A three-parameter asymmetric Laplace distribution and its extension. Commun Stat Theory Methods2005; 34: 1867–1879.
MarinoMFAlfóM. Gaussian quadrature approximations in mixed hidden Markov models for longitudinal data: a simulation study. Comput Stat Data Anal2016; 94: 193–209.
25.
LairdN. Nonparametric maximum likelihood estimation of a mixing distribution. J Am Stat Assoc1978; 73: 805–811.
26.
DempsterALairdNMRubinDB. Maximum likelihood from incomplete data via the EM algorithm. J R Stat Soc B1977; 39: 1–38.
27.
AitkinM. A general maximum likelihood analysis of variance components in generalized linear models. Biometrics1999; 55: 117–128.
28.
DemirtasHSchaferJL. On the performance of random-coefficient pattern-mixture models for non-ignorable drop-out. Stat Med2003; 22: 2553–2575.
29.
LittleRJARubinDB. Statistical analysis with missing data, Hoboken, NJ: John Wiley & Sons, 2002.
30.
HeckmanJJ. The common structure of statistical models of truncation, sample selection and limited dependent variables and a simple estimator for such models. Ann Econ Social Meas1976; 5: 475–492.
31.
LittleRJA. Pattern-mixture models for multivariate incomplete data. J Am Stat Assoc1993; 88: 125–134.
32.
WuMCCarrollRJ. Estimation and comparison of changes in the presence of informative right censoring by modeling the censoring process. Biometrics1988; 44: 175–188.
33.
WulfsohnMSTsiatisAA. A joint model for survival and longitudinal data measured with error. Biometrics1997; 53: 330–339.
34.
LittleRJA. Modeling the drop-out mechanism in repeated-measures studies. J Am Stat Assoc1995; 90: 1112–1121.
35.
RizopoulosDLesaffreE. Introduction to the special issue on joint modelling techniques. Stat Meth Med Res2014; 23: 3–10.
36.
BartolucciFFarcomeniA. A discrete time event-history approach to informative drop-out in mixed latent Markov models with covariates. Biometrics2015; 71: 80–89.
37.
MaruottiA. Handling non-ignorable dropouts in longitudinal data: a conditional model based on a latent Markov heterogeneity structure. TEST2015; 24: 84–109.
38.
DantanEProust-LimaCLetenneurLet al.Pattern mixture models and latent class models for the analysis of multivariate longitudinal data with informative dropouts. Int J Biostat2008; 4: 1–26.
39.
MaGTroxelABHeitjanDF. An index of local sensitivity to nonignorable drop-out in longitudinal modelling. Stat Med2005; 24: 2129–2150.
40.
AgrestiA. Analysis of ordinal categorical data, Hoboken, NJ: John Wiley & Sons, 2010.
41.
BaumLEPetrieTSoulesGet al.A maximization technique occurring in the statistical analysis of probabilistic functions of Markov chains. Ann Math Stat1970; 41: 164–171.
42.
WelchLR. Hidden Markov models and the Baum-Welch algorithm. IEEE Inform Theory Society Newsletter2003; 53: 10–13.
43.
AkaikeHInformation theory and an extension of the maximum likelihood principle. In: PetrovBNCsakiF (eds). Second International symposium on information theory, Budapest: Akadémiai Kiado, 1973, pp. 267–281.
44.
SchwarzG. Estimating the dimension of a model. Ann Stat1978; 6: 461–464.
45.
JonesRH. Bayesian information criterion for longitudinal and clustered data. Stat Med2011; 30: 3050–3056.
46.
LahiriS. Theoretical comparisons of block bootstrap methods. Ann Stat1999; 27: 386–404.
47.
BuchinskyM. Estimating the asymptotic covariance matrix for quantile regression models. A Monte Carlo study. J Econom1995; 68: 303–338.
48.
RadloffLS. The CES-D scale a self-report depression scale for research in the general population. Appl Psychol Meas1977; 1: 385–401.
49.
RandWM. Objective criteria for the evaluation of clustering methods. J Am Stat Assoc1971; 66: 846–850.
Supplementary Material
Please find the following supplemental material available below.
For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.
For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.
