PENANGANAN OVERDISPERSI PADA MODEL REGRESI POISSON MENGGUNAKAN MODEL REGRESI BINOMIAL NEGATIF
Poisson regression is the most popular tool for modeling the relationship between a discrete data in the response variable and a set of predictors with continue, discrete, categoric or mix data. Response variable with discrete data, however, may overdispersed or underdispersed, not conductive to Poi...
| Main Authors: | , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Universitas Diponegoro
2011-12-01
|
| Series: | Media Statistika |
| Online Access: | https://ejournal.undip.ac.id/index.php/media_statistika/article/view/2470 |
| Summary: | Poisson regression is the most popular tool for modeling the relationship between a discrete data in the response variable and a set of predictors with continue, discrete, categoric or mix data. Response variable with discrete data, however, may overdispersed or underdispersed, not conductive to Poisson regression which assumed that the mean value equals to variance (equidispersed). One of the model that be used to overdispersed the discrete data is a regression model based on mixture distribution namely Poisson-gamma mixture which result negative binomial distribution. This regression model usually known as binomial negative regression. Using Generalized Linier Model (GLM) approach, the given model, parameter estimate, diagnostics, and interpretation of negative binomial regression can be determined.
Keyword: Negative Binomial Distribution, Dispersion, Generalized Linier Model |
|---|---|
| ISSN: | 1979-3693 2477-0647 |