A Bayesian test of homogeneity of association parameter using transition modelling of longitudinal mixed responses. Issue 10 (26th July 2016)

• Record Type: Journal Article
• Authors: Noorian, S.; Ganjali, M.; Bahrami Samani, E.
• Is Part Of: Journal of applied statistics. Volume 43:Issue 10(2016)
• Page Start: 1850
• Page End: 1863
• Publication Date: 2016-07-26
• Subjects: Bayesian analysis -- joint modelling -- longitudinal studies -- mixed ordinal and continuousresponses -- PIAT data -- transition model; 62P99; Statistics -- Periodicals; 519.5
• DOI: 10.1080/02664763.2015.1125858
• Languages: English
• ISSNs: 0266-4763

Abstract

ABSTRACT: In this paper, a Bayesian framework using a joint transition model for analysing longitudinal mixed ordinal and continuous responses is considered. The joint model considers a multivariate mixed model for the responses in which a transitive cumulative logistic regression model and an autoregressive regression model are used to model ordinal and continuous responses, respectively. Also, to take into account the association between longitudinal ordinal and continuous responses, a dynamic association parameter is used. A test is conducted to see whether this parameter is time-invariant and another test is presented to see whether this parameter is equal to zero or significantly far from zero. Our approach is applied to longitudinal PIAT (Peabody Individual Achievement Test) data where the Bayesian estimates of parameters are obtained.