Abstract:
The objectives of this thesis are to propose a new mixed negative binomial distribution, to derive parameter estimation of the proposed distribution and to compare efficary of the proposed distribution with other distributions for count data analysis. The results of this research show that the negative binomial two parameter weighted exponential distribution is the new mixed negative binomial distribution obtained by mixing the negative binomial distribution with the two parameter weighted exponential distribution. There are some special models, e.g. the negative binomial exponential, negative binomial gamma, and negative binomial weighted exponential distributions are presented as special cases of this negative binomial two parameter weighted exponential distribution.
The parameter estimation of the negative binomial two parameter weighted exponential distribution is estimated by maximum likelihood estimation and Bayesian approach. Most of simulation results, we can see that the Bayesian method likely outperforms the maximum likelihood method. In application studies, the goodness of fit test based on the Poisson, negative binomial and negative binomial two parameter weighted exponential distribution are discussed. The results emphasize that the negative binomial two parameter weighted exponential distribution has a better fit than the other competitive distributions. Therefore, the negative binomial two parameter weighted exponential distribution is alternative choices for data that has an overdispersion problem and heavy-tailed with positive skewness.
Kasetsart University. Office of the University Library