Analyzing Multivariate Count Data Using a GQL Based on Copula Structures Naushad Mamode Khan (1) (1) , Yuvraj Sunecher (2) and Vandna Jowaheer (3) University of Mauritius, Reduit, Mauritius, [email protected] (2) University of Mauritius, Reduit, Mauritius, [email protected] (3) University of Mauritius, Reduit, Mauritius, [email protected] Abstract In the last few years, the modelling of multivariate count data has been a topic of concern for many researchers in the field of epidemiology, agriculture, economics and finance. The most recent findings in the analysis of such data illustrate that it is easier to specify the likelihood function of these multivariate count responses through the use of copula constructors such as Clayton and Frank copulas[1, 2]. However, in a regression set-up, the resulting maximum likelihood estimation procedure involves huge computationally intensive expressions that make the methodology almost unfeasible. This raises the need to explore some other parsimonious estimation methodologies. In this context, this paper proposes an alternative estimation scheme based on the generalized quasi-likelihood (GQL) approach to estimate the regression and dispersion coefficients. The performance of the GQL method is assessed through simulation studies based on generating Poisson and Negative Binomial counts that are exposed to some covariate patterns. Keywords: Counts, multivariate, copula, GQL. References  A.K. Nikoloulopoulos and D. Karlis, Modelling multivariate count data using copulas, Communications in Statistics, Simulation and Computation, 2010.  A.K. Nikoloulopoulos and D. Karlis, Regression in a copula model for bivariate count data, Journal of Applied Statistics, 2010.
© Copyright 2021