Longitudinal zero-inflated count data arise frequently in substance use research when assessing the effects of behavioral and pharmacological interventions. Zero-inflated count models (e.g. zero-inflated Poisson or zero-inflated negative binomial) with random effects have been developed to analyze this type of data. In random effects zero-inflated count models, the random effects covariance matrix is typically assumed to be homogeneous (constant across subjects). However, in many situations this matrix may be heterogeneous (differ by measured covariates). In this paper, we extend zero-inflated count models to account for random effects heterogeneity by modeling their variance as a function of covariates. We show via simulation that ignoring intervention and covariate-specific heterogeneity can produce biased estimates of covariate and random effect estimates. Moreover, those biased estimates can be rectified by correctly modeling the random effects covariance structure. The methodological development is motivated by and applied to the Combined Pharmacotherapies and Behavioral Interventions for Alcohol Dependence (COMBINE) study, the largest clinical trial of alcohol dependence performed in United States with 1383 individuals.
MaruottiA. A two-part mixed-effects pattern-mixture model to handle zero-inflation and incompleteness in a longitudinal setting. Biometrical J2011; 53(5): 716–734.
2.
DeSantisSMLazaridisCJiS, et al.Analyzing propensity matched zero-inflated count outcomes in observational studies. J Appl Statist2013; 127: 1–15.
3.
Hur K. A random effects zero-inflated Poisson regression model for clustered extra-zero count data. PhD dissertation, University of Illinois at Chicago, 1999.
4.
BuuRTXLiAZuckerRA. Statistical models for longitudinal zero-inflated count data with applications to the substance abuse field. Statist Med2012; 31: 4074–4086.
LambertD. Zero-inflated Poisson regression, with an application to defects in manufacturing. Technometrics1992; 34: 1–14.
7.
RidoutMHindeJDemeAtrioCG. A score test for testing a zero-inflated poisson regression model against zero-inflated negative binomial alternatives. Biometrics2001; 57(1): 219–223.
8.
HallDB. Zero-inflated Poisson and binomial regression with random effects: a case study. Biometrics2000; 56(4): 1030–1039.
9.
MinYAgrestiA. Random effect models for repeated measures of zero-inflated count data. Statist Modell2005; 5(1): 1–19.
10.
LeeAHWangKScottJA, et al.Multi-level zero-inflated Poisson regression modelling of correlated count data with excess zeros. Statist Meth Med Res2006; 15(1): 47–61.
11.
DagneGA. Bayesian semiparametric zero-inflated poisson model for longitudinal count data. Math Biosci2010; 224(2): 126–130.
12.
NeelonBHMalleyAJNormandSLT. A Bayesian model for repeated measures zero-inflated count data with application to outpatient psychiatric service use. Statist Modell2010; 10(4): 421–439.
13.
PourahmadiMDanielsM. Dynamic conditionally linear mixed models for longitudinal data. Biometrics2002; 58(1): 225–231.
14.
DanielsMJZhaoYD. Modelling the random effects covariance matrix in longitudinal data. Statist Med2003; 22(10): 1631–1647.
15.
GueorguievaRWuRDonovanD, et al.Naltrexone and combined behavioral intervention effects on trajectories of drinking in the COMBINE study. Drug Alcohol Depend2010; 107: 221–229.
16.
WitkiewitzKMaistoSADonovanDM. A comparison of methods for estimating change in drinking following alcohol treatment. Alcoholism2010; 34(12): 2116–2125.
17.
HeagertyPJKurlandBF. Misspecified maximum likelihood estimates and generalised linear mixed models. Biometrika2001; 88(4): 973–985.
18.
ChiuTYLeonardTTsuiKW. The matrix-logarithmic covariance model. J Am Statist Assoc1996; 91(433): 198–210.
19.
DavidianM. Nonlinear Models for Repeated Measurement Data(Chapman & Hall/CRC Monographs on Statistics and Applied Probability, vol. 62). Boca Raton, FL: CRC Press, 1995.
20.
LinXRazJHarlowSD, et al.Linear mixed models with heterogeneous within-cluster variances. Biometrics1997; 53(3): 910–923.
21.
ZhangFWeissRE. Diagnosing explainable heterogeneity of variance in random-effects models. Can J Statist2000; 28(1): 3–18.
22.
DanielsMJKassRE. Nonconjugate Bayesian estimation of covariance matrices and its use in hierarchical models. J Am Statist Assoc1999; 94(448): 1254–1263.
23.
DanielsMJKassRE. Shrinkage estimators for covariance matrices. Biometrics2001; 57(4): 1173–1184.
24.
HuangXLiGElashoffRM. A joint model of longitudinal and competing risks survival data with heterogeneous random effects and outlying longitudinal measurements. Statist Inference2010; 3(2): 185–195.
25.
PourahmadiM. Joint mean-covariance models with applications to longitudinal data: Unconstrained parameterisation. Biometrika1999; 86(3): 677–690.
26.
HuangXLiGElashoffRM, et al.A general joint model for longitudinal measurements and competing risks survival data with heterogeneous random effects. Lifetime Data Anal2011; 17(1): 80–100.
27.
AntonRFO’MalleySSCirauloDA, et al.Combined pharmacotherapies and behavioral interventions for alcohol dependence: the combine study: a randomized controlled trial. JAMA2006; 295(17): 2003–2017.
28.
CarmenBAngelesMAnaM, et al.Efficacy and safety of naltrexone and acamprosate in the treatment of alcohol dependence: a systematic review. Addiction2004; 99(7): 811–828.
29.
SobellLCBrownJLeoGI, et al.The reliability of the alcohol timeline followback when administered by telephone and by computer. Drug Alcohol Depend1996; 42(1): 49–54.
30.
SobellLSobellM. Handbook of Psychiatric Measures, Washington, DC: American Psychiatric Association, 2000, pp. 477–479.
31.
DayAMCelioMALismanSA, et al.Gender, history of alcohol use and number of drinks consumed predict craving among drinkers in a field setting. Addictive Beh2014; 39(1): 354–357.
32.
RobertsSC. Whether men or women are responsible for size of gender gap in alcohol consumption depends on alcohol measure: A study across U.S. states. Contemp Drug Problems2012; 39: 195–212.
33.
WilsnackRWWilsnackSCKristjansonAF, et al.Gender and alcohol consumption: patterns from the multinational Genacis project. Addiction2009; 104(9): 1487–1500.
34.
BaeSFamoyeFWuluJT, et al.A rich family of generalized poisson regression models with applications. Math Comput Sim2005; 69(1-2): 4–11.
35.
LoquihaOHensNChavaneL, et al.Modeling heterogeneity for count data: a study of maternal mortality in health facilities in Mozambique. Biometrical J2013; 55(5): 647–660.
36.
LesaffreESpiessensB. On the effect of the number of quadrature points in a logistic random effects model: an example. J R Statist Soc: Ser C2001; 50(3): 325–335.
37.
AbramowitzMStegunIA. Handbook of Mathematical Functions: with Formulas, Graphs, and Mathematical Tables(Dover Books on Mathematics). New York: Dover Publications, 1965.
38.
GolubGHWelschJH. Calculation of Gauss Quadrature Rules, Providence, RI: American Mathematical Society, 1969.
39.
SAS Version 9.2, 2008.
40.
Akaike H. Information theory and an extension of the maximum likelihood principle. In Second International Symposium on Information Theory, 1973, pp. 267–281.
41.
SchwarzG. Estimating the dimension of a model. Ann Statist1978; 6(2): 461–464.
42.
WangJXieHFisherJH. Multilevel Models. Applications using SAS, De Gruyter, 2011.
43.
SelfSGLiangKY. Asymptotic properties of maximum likelihood estimators and likelihood ratio tests under nonstandard conditions. J Am Statist Assoc1987; 82(398): 605–610.
44.
SwiftRPettinatiHM. Choosing pharmacotherapies for the combine study–process and procedures: an investigational approach to combination pharmacotherapy for the treatment of alcohol dependence. J Stud Alcohol Drugs2005; 15: 141–141.
45.
DeSantisSMBandyopadhyayDBakerNL, et al.Modeling longitudinal drinking data in clinical trials: an application to the COMBINE study. Drug Alcohol Depend2013; 132: 244–250.
46.
McCullochCENeuhausJM. Prediction of random effects in linear and generalized linear models under model misspecification. Biometrics2011; 67(1): 270–279.
47.
McCullochCENeuhausJM, et al.Misspecifying the shape of a random effects distribution: why getting it wrong may not matter. Statist Sci2011; 26(3): 388–402.
48.
VerbekeGLesaffreE. A linear mixed-effects model with heterogeneity in the random-effects population. J Am Statist Assoc1996; 91(433): 217–221.
49.
GhideyWLesaffreEEilersP. Smooth random effects distribution in a linear mixed model. Biometrics2004; 60(4): 945–953.
50.
ZhangDDavidianM. Linear mixed models with flexible distributions of random effects for longitudinal data. Biometrics2001; 57(3): 795–802.
51.
XieFCWeiBCLinJG. Score tests for zero-inflated generalized Poisson mixed regression models. Computat Statist Data Anal2009; 53(9): 3478–3489.
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.
