A joint modeling approach to data with informative cluster size: Robustness to the cluster size model

Chen, Zhen; Zhang, Bo; Albert, Paul S.

doi:10.1002/sim.4239

Cited by 24 publications

(38 citation statements)

References 23 publications

(29 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The N K S B P N I , N K S B P I − N , and c N K S B P I models provided ER curves that agreed less with the truth, though the fit was similar at the lower doses. We note that using an incorrect distribution assumption for the cluster size model may induce biases (Chen et al ), which is supported by the results that the N K S B P I − N model induced a considerable bias in the ER curves.…”

Section: Simulation Studiessupporting

confidence: 79%

“…In addition, the distribution of the cluster‐specific random effect b i is assigned a nonparametric DP prior with precision α b and base parametric distribution F b 0 as follows:

b_{i} \sim F_{b} = true \sum_{s = 1}^{\infty} q_{s} δ_{θ_{s}}, q_{s} = W_{s} true \prod_{l = 1}^{s - 1} false(1 - W_{l} false), W_{s} \sim Beta false(1, α_{b} false), θ_{s} \sim F_{b 0},

where δ θ is a point mass at θ and all W s 's and θ s 's are independent. Note that the DP assigned to F b relaxes the distribution assumption of the cluster size model, which is important in light of the work of Chen et al (). Our model imposes two different latent groupings of the clusters (i.e., litters): The nested KSBP allows a group of clusters to share the same distribution for the subject‐specific random effects (the U i j 's) and the DP allows a group of clusters to share the same latent covariate value b i (e.g., certain dams may share the same unmeasured pregnancy complication, which could affect litter size and fetal outcomes).…”

Section: Methodsmentioning

confidence: 99%

“…For instance, the number of pups per litter often decreases with increasing toxin levels and fetal weight decreases with increasing litter size. Failure to account for these types of relationships may induce biases in models (Chen, Zhang, & Albert, 2011;Dunson, Chen, & Harry, 2003;Gueorguieva, 2005). For instance, if the true causal effect of a toxin is to decrease both fetal weight and litter size, an analysis that ignores the relationship between these outcomes could underestimate the inverse relationship between dose and fetal weight (Romero, Villamayor, Grau, Sacristan, & Ortiz, 1992).…”

mentioning

confidence: 99%

See 2 more Smart Citations

Semiparametric Bayesian joint modeling of clustered binary and continuous outcomes with informative cluster size in developmental toxicity assessment

Hwang

Pennell

2018

Environmetrics

View full text Add to dashboard Cite

Many authors have shown that joint modeling outcomes can effectively improve statistical power in studies involving clustered data. For instance, in developmental toxicity studies, fetal weight and malformation have been jointly analyzed to obtain more efficient inferences. Unfortunately, in these studies, the number of subjects is often correlated with the outcome values, which can result in biased estimates if the relationship is not accounted for. Most methods for joint modeling data with informative cluster size assume standard parametric response distributions. However, in developmental toxicity studies, it is possible that response distributions vary in location and shape with dose, which may not be easily captured by parametric models. Motivated by applications in developmental toxicity, we propose a semiparametric Bayesian joint model for clustered binary and continuous responses with informative cluster size. In our model, a nested kernel stick‐breaking process prior is assigned to the distribution of a random effect shared across outcomes, which allows simultaneous grouping of clusters and subjects within clusters, and flexible changes in distribution shape with dose shared across outcomes. We account for informative cluster size by modeling the number of subjects in each cluster using a Poisson regression model with a cluster‐specific random effect that is shared with the outcome variables. The Poisson assumption is relaxed by assigning a Dirichlet process prior to the unknown distribution of the cluster‐specific random effect. We apply our method to data from a developmental toxicity study of diethylhexyl phthalate.

show abstract

Section: Simulation Studiessupporting

confidence: 79%

“…In addition, the distribution of the cluster‐specific random effect b i is assigned a nonparametric DP prior with precision α b and base parametric distribution F b 0 as follows:

b_{i} \sim F_{b} = true \sum_{s = 1}^{\infty} q_{s} δ_{θ_{s}}, q_{s} = W_{s} true \prod_{l = 1}^{s - 1} false(1 - W_{l} false), W_{s} \sim Beta false(1, α_{b} false), θ_{s} \sim F_{b 0},

Section: Methodsmentioning

confidence: 99%

mentioning

confidence: 99%

See 1 more Smart Citation

Semiparametric Bayesian joint modeling of clustered binary and continuous outcomes with informative cluster size in developmental toxicity assessment

Hwang

Pennell

2018

Environmetrics

View full text Add to dashboard Cite

show abstract

“…Explicitly creating a model for the intervisit times and incorporating it in a model for the outcomes of interest would be especially difficult given the “spells” of concentrated visits. We note that Chen et al showed that misspecification of the joint distribution of the outcome and number of visits can produce biased estimation.…”

Section: Introductionmentioning

confidence: 84%

Analysis of longitudinal data from outcome‐dependent visit processes: Failure of proposed methods in realistic settings and potential improvements

Neuhaus

McCulloch

Boylan

2018

Statistics in Medicine

View full text Add to dashboard Cite

The timing and frequency of the measurement of longitudinal outcomes in databases may be associated with the value of the outcome. Such visit processes are termed outcome dependent, and previous work showed that conducting standard analyses that ignore outcome-dependent visit times can produce highly biased estimates of the associations of covariates with outcomes. The literature contains several classes of approaches to analyze longitudinal data subject to outcome-dependent visit times, and all of these are based on simplifying assumptions about the visit process. Based on extensive discussions with subject matter investigators, we identified common characteristics of outcome-dependent visit processes that allowed us to evaluate the performance of existing methods in settings with more realistic visit processes than have been previously investigated. This paper uses the analysis of data from a study of kidney function, theory, and simulation studies to examine a range of settings that vary from those where all visits have a low degree of missingness and outcome dependence (which we call "regular" visits) to those where all visits have a high degree of missingness and outcome dependence (which we call "irregular" visits). Our results show that while all the approaches we studied can yield biased estimates of some covariate effects, other covariate effects can be estimated with little bias. In particular, mixed effects models fit by maximum likelihood yielded little bias in estimates of the effects of covariates not associated with the random effects and small bias in estimates of the effects of covariates associated with the random effects. Other approaches produced estimates with greater bias. Our results also show that the presence of some regular visits in the data set protects mixed model analyses from bias but not other methods.

show abstract

“…Chen et al . () further studied the robustness of the SRPM under model misspecification. Seaman et al .…”

Section: Introductionmentioning

confidence: 97%

Pattern–Mixture Models with Incomplete Informative Cluster Size: Application to a Repeated Pregnancy Study

Chaurasia

Liu

Albert

2017

Journal of the Royal Statistical Society Series C: Applied Statistics

Self Cite

View full text Add to dashboard Cite

Summary. The incomplete informative cluster size problem is motivated by the National Institute of Child Health and Human Development consecutive pregnancies study, aiming to study the relationship between pregnancy outcomes and parity. These pregnancy outcomes are potentially associated with the number of births over a woman's lifetime, resulting in an incomplete informative cluster size (censored at the end of the study window).We develop a pattern-mixture model for informative cluster size by treating the lifetime number of births as a latent variable. We compare this approach with a simple alternative method that approximates the pattern-mixture model. We show that the latent variable approach has good statistical properties for estimating both the mean trajectory of birth weight and the proportion of gestational hypertension with increasing parity.

show abstract

A joint modeling approach to data with informative cluster size: Robustness to the cluster size model

Cited by 24 publications

References 23 publications

Semiparametric Bayesian joint modeling of clustered binary and continuous outcomes with informative cluster size in developmental toxicity assessment

Semiparametric Bayesian joint modeling of clustered binary and continuous outcomes with informative cluster size in developmental toxicity assessment

Analysis of longitudinal data from outcome‐dependent visit processes: Failure of proposed methods in realistic settings and potential improvements

Pattern–Mixture Models with Incomplete Informative Cluster Size: Application to a Repeated Pregnancy Study

Contact Info

Product

Resources

About