MODEL REGRESI DIAGONAL INFLATED BIVARIATE POISSON PADA DATA OLAHRAGA; DIAGONAL INFLATED BIVARIATE POISSON REGRESSION MODEL IN SPORT DATA

ARISDYAN PRASETYA, Zulaela

2015 | Skripsi | FMIPA

Bivariate Poisson models appear in a broad context in the calculation of the data pairs. In addition to the sports field, this model can also be applied in various other fields like marketing, epidemiology, health and econometrics and still many other fields. Bivariate Poisson distribution model is used to model thesports data especially football. Independent Poisson distribution is usually used to model the number of goals of two teams competing in a game. One problem that arises is bivariate Poisson models only model the data so as to be positively correlated but for negatively correlated data still can not be addressed by this model. Besides the Poison distribution properties can not be modeled data that is overdisperssion or underdisperssion. Overdisperssion means that the variance value exceeds the mean value as well as for underdispersi which means that the variance value is smaller than the mean value. Therefore considered a model mixture of bivariate Poisson models, namely the model diagonal bivariate inflated Poisson (DIBP). DIBP models can also easily to computing. Furthermore, this thesis will discuss the process of estimation and model building DIBP using EM algorithm.

Kata Kunci : Overdispersion; Diagonal inflated; Bivariate Poisson Regression; Sport data