作者: Sonya K. Sterba
DOI: 10.1080/10705511.2016.1250635
关键词:
摘要: Many psychology applications assess measurement invariance of a construct (e.g., depression) over time. These are often characterized by few time points 3), but high rates dropout. Although such routinely assume that the dropout mechanism is ignorable, this assumption may not always be reasonable. In presence nonignorable dropout, fitting conventional longitudinal factor model (LFM) to can yield misleading inferences about level invariance, along with biased parameter estimates. article we develop pattern mixture models (PM-LFMs) for quantifying uncertainty in testing due an unknown, potentially nonignorable, mechanism. PM-LFMs kind multiple group wherein observed missingness patterns define groups, LFM parameters differ across these pattern-groups subject identification constraints, and marginal inference longitudinal...