PEMODELAN REGRESI ZERO-INFLATED NEGATIVE BINOMIAL (ZINB) UNTUK DATA RESPON DISKRIT DENGAN EXCESS ZEROS

Bayu Ariawan, Suparti Suparti, Sudarno Sudarno

Abstract


Zero-Inflated Negative Binomial (ZINB) regression is one of the methods used in troubleshooting overdispersion due to excessive zero values ​​in the response variable (excess zeros). ZINB regression model was based on the negative binomial distribution resulting from a mixture distribution between Poisson distribution  withis value of random variable which gamma distributed.

ZINB regression parameter estimation can be performed by using Maximum Likelihood Estimation (MLE) method then is followed by the EM algorithm (Expectation maximization) procedure and Newton Rhapson. Test the suitability of the model simultaneously performed using Likelihood Ratio test and significance testing parameters individually performed with Wald test statistics. The model is applied to the case of car insurance obtained PT. Insurance of Sinar Mas Semarang Branch in 2010 in the form of data many policyholders filed claims to the PT. Sinar Mas Semarang Branch Insurance. Response variable is the number of claims submitted to the PT. Insurance of Sinar Mas Semarang Branch, while the predictor  variable is the age car and the type of coverage that consists of All Risk, Total Lost Only (TLO), and the joint between All Risk and Total Lost Only (TLO). From the analytical result obtained the conclution that the age of the car and the type of coverage affects number of claims filed by the policyholder to the PT. Insurance of Sinar Mas Semarang Branch in 2010.


Keywords


Overdispersion, Excess zeros, Negative Binomial Distribution, Zero-Inflated Negative Binomial (ZINB) Regression

Full Text:

PDF

Refbacks

  • There are currently no refbacks.



Creative Commons License
Jurnal Gaussian by Departemen Statistika Undip is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.

Flag Counter