Innovative Applications of Shared Random Parameter Models for Analyzing Longitudinal Data Subject to Dropout
- 968 Downloads
Shared random parameter (SRP) models provide a framework for analyzing longitudinal data with missingness. We discuss the basic framework and review the most relevant literature for the case of a single outcome followed longitudinally. We discuss estimation approaches, including an approximate approach which is relatively simple to implement. We then discuss three applications of this framework in novel settings. First, we show how SRP models can be used to make inference about pooled or batched longitudinal data subject to non-ignorable dropout. Second, we show how one of the estimation approaches can be extended for estimating high dimensional longitudinal data subject to dropout. Third, we show how to use jointly model complex menstrual cycle length data and time to pregnancy in order to study the evolution of menstrual cycle length accounting for non-ignorable dropout due to becoming pregnant and to develop a predictor of time-to-pregnancy from repeated menstrual cycle length measurements. These three examples demonstrate the richness of this class of models in applications.
KeywordsLongitudinal Measurement Menstrual Cycle Length Random Effect Distribution Pattern Mixture Model Informative Dropout
We thank the referee and editor for their thoughtful and constructive comments which lead to an improved manuscript. We also thank the audience of the International Symposium in Statistics (ISS) on Longitudinal Data Analysis Subject to Outliers, Measurement Errors, and/or Missing Values. This research was supported by the Intramural Research Program of the Eunice Kennedy Shriver National Institute of Child Health and Human Development.
- Albert, P.S., Shih, J.H.: Modelling batched Gaussian longitudinal weight data in mice subject to informative dropout. Published on-line before print. Stat. Methods Med. Res. (2011)Google Scholar
- Guo, Y., Manatunga, A.K., Chen, S., Marcus, M.: Modeling menstrual cycle length using a mixture distribution. Biostatistics 7, 151–168 (2006)Google Scholar
- Molenberghs, G., Njeru, E., Kenward, M.G., Verbeke, G.: Enriched-data problems and essential non-identifiability. Int. J. Stat. Med. Res. 1, 16–44 (2012)Google Scholar
- Tsiaits, A.A., Davidian, M.: Joint modeling of longitudinal and time-to-event data: an overview. Stat. Sin. 14, 809–834 (2004)Google Scholar
- Tsiatis, A.A., DeGruttola, V., Wulfsohn, M.S.: Modeling the relationship of survival to longitudinal data measured with error. Applications to survival and CD4 counts in patients with AIDS. J. Am. Stat. Assoc. 90, 27–37 (1995)Google Scholar