|
Comparing Marginal and Transition Models In The Analysis of Binary Longitudinal Data: a Simulation Study
|
| F Zayeri, S Shahsavari, Ar Baghestani, S Jambarsang *1, V Lohrabian |
| 1- , s.jambarsang@sbmu.ac.ir |
|
|
Abstract: (10380 Views) |
| Introduction: Longitudinal studies widely used in many branches of science, especial-lly medical science. In the past two deca-des, many models have been proposed for the evaluation of such data. These models are also developed to investigate various re-sponse variables such that categorical varia-bles that have many applications in medical research. While many papers have comp-ared the marginal and random effects mo-dels but fewer studies have compared the marginal and transition models.
Materials & Methods: This paper used simulation with different scenarios for corr-elation, sample size and repeated measures and calculated ROC Curve for two models. Then estimated AUC for them and comp-arison was performed.
Findings: If intensity transition for healthy to disease was large and patient remained disease for a long time that is better used transition model to prediction. In addition, in conditions that intensity transition in any status was equal or trends of transition inte-nsity was inverse of position1 that healthy people remained health for a long time and patients recovered quickly, that is better used marginal model.
Discussion & Conclusion: In analysis of longitudinal data for achieving more accur-ate results when purpose of study was pred-iction, that is better intensity transition bet-ween status was considered and then dec-ided which model is choosen. |
|
| Keywords: binary data, markov correlation, marginal model, transition model, roc area under curve |
|
|
Full-Text [PDF 723 kb]
(3072 Downloads)
|
Type of Study: Research |
Subject:
biostatistics
Received: 2013/03/3 | Accepted: 2013/03/17 | Published: 2013/03/17
|
|
|
|
|
|
|
| Send email to the article author |
|
|
|
| Add your comments about this article |
|
|
|
