A conditional count model for repeated count data and its application to GEE approach
In this article, a conditional model is proposed for modeling longitudinal count data. The joint density is disintegrated into the marginal and conditional densities according to the multiplication rule. It allows both positive and negative correlation among variables, which most multivariate count models do not possess. To show the efficiency of the proposed model for count data, we have applied to the generalized estimating equations and the inverse Fisher information matrix is compared with the covariance matrix from estimating equations. A simulation experiment is displayed and an application of the model to divorce data is presented. In addition, a comparison of conditional model and bivariate Poisson model proposed by Kocherlakota and Kocherlakota has shown using simulated data.
KeywordsCount data Generalized estimating equations Conditional probability
We are grateful to the HRS (Health and Retirement Study) which is sponsored by the National Institute of Aging (Grant Number NIA U01AG09740) and conducted by the University of Michigan.
- Health and Retirement Study (Wave [1-7]/Year [1992–2004]), Public use dataset, produced and distributed by the University of Michigan with funding from National Institute on Aging. Ann. Arbor, MIGoogle Scholar
- Kocherlatoka S, Kocherlakota K (1993) Bivariate discrete distributions. Marcel Dekker, New YorkGoogle Scholar
- Minka TP, Shmueli G, Kadane JB, Borle S, Boatwright P (2003) Computing with the COM- Poisson distribution. Technical report 776, Department of Statistics, Carnegie Mellon University. http://www.stat.cmu.edu/tr/tr776/tr776.html
- Nikoloulopoulos AK (2015) Efficient estimation of high-dimensional multivariate normal copula models with discrete spatial responses. Stoch Environ Res Risk Assess:1–13 doi: 10.1007/s00477-015-1060-2