Bias in 2-part mixed models for longitudinal semicontinuous data

Mixed model Censoring (clinical trials) Independence
DOI: 10.1093/biostatistics/kxn044 Publication Date: 2009-01-10T01:12:38Z
ABSTRACT
Semicontinuous data in the form of a mixture zeros and continuously distributed positive values frequently arise biomedical research. Two-part mixed models with correlated random effects are an attractive approach to characterize complex structure longitudinal semicontinuous data. In practice, however, independence assumption about these may often be made for convenience computational feasibility. this article, we show that bias can induced regression coefficients when truly but misspecified as independent 2-part model. Paralleling work on under nonignorable missingness within shared parameter model, derive investigate asymptotic selected settings models. The performance practice is further evaluated using Monte Carlo simulations. Additionally, potential investigated artificial zeros, due left censoring from some detection or measuring limit, incorporated. To illustrate, fit different University Toronto Psoriatic Arthritis Clinic, aim being examine whether there differential disease activity damage physical functioning measured by health assessment questionnaire scores over course psoriatic arthritis. Some practical issues variance component estimation revealed through analysis considered.
SUPPLEMENTAL MATERIAL
Coming soon ....
REFERENCES (25)
CITATIONS (82)
EXTERNAL LINKS
PlumX Metrics
RECOMMENDATIONS
FAIR ASSESSMENT
Coming soon ....
JUPYTER LAB
Coming soon ....