2017/07/07 by Alessandra Spagnoli, Spagnoli, Alessandra, Maria Francesca Marino +3
Computer Science · Mathematics · Medicine · #Bayesian Methods and Mixture Models #FOS: Computer and information sciences #Mathematical and Theoretical Epidemiology and Ecology Models #Methodology (stat.ME) #Statistical Methods and Bayesian Inference
paper · pdf · doi:10.48550/arxiv.1707.02182
openalex publication_date 2017/07/07 · openalex created_date 2022/08/09 · openalex updated_date 2026/07/28
In longitudinal studies, subjects may be lost to follow-up, or miss some of\nthe planned visits, leading to incomplete response sequences. When the\nprobability of non-response, conditional on the available covariates and the\nobserved responses, still depends on unobserved outcomes, the dropout mechanism\nis said to be non ignorable. A common objective is to build a reliable\nassociation structure to account for dependence between the longitudinal and\nthe dropout processes. Starting from the existing literature, we introduce a\nrandom coefficient based dropout model where the association between outcomes\nis modeled through discrete latent effects. These effects are outcome-specific\nand account for heterogeneity in the univariate profiles. Dependence between\nprofiles is introduced by using a bi-dimensional representation for the\ncorresponding distribution. In this way, we define a flexible latent class\nstructure which allows to efficiently describe both dependence within the two\nmargins of interest and dependence between them. By using this representation\nwe show that, unlike standard (unidimensional) finite mixture models, the non\nignorable dropout model properly nests its ignorable counterpart. We detail the\nproposed modeling approach by analyzing data from a longitudinal study on the\ndynamics of cognitive functioning in the elderly. Further, the effects of\nassumptions about non ignorability of the dropout process on model parameter\nestimates are (locally) investigated using the index of (local) sensitivity to\nnon-ignorability.\n