Poisson regression is frequently used to model count data. It requires an assumption that the mean of response variable must be equal to its variance (equidispersion). However, this equidispersion condition rarely occurs. Mostly, the variance of the response variable is higher than its mean (overdispersion). One of the causes of overdispersion is excess zeros. Zero-Inflated Poisson (ZIP) Regression provides the solution to model count data with excess zeros. This model can handle overdispersion since the variance of ZIP Distribution is higher than its mean. The assumptions of response variable are the only zero observations occurred with probability p, otherwise has Poisson distribution with probability 1-p. We then apply the model to the number of maternal pregnant deaths cases in Depok, Bogor, and Sukabumi in 2015. The data was provided by Indonesian Department of Health. The ZIP Regression model shows that sum of maternal visit to community health clinic and sum of 90 Fe tablet compliances affects significantly to the number of maternal pregnant deaths.
|Publication status||Published - 2018|
|Event||The 8 th Annual Basic Science International Conference - ID, Malang, Indonesia|
Duration: 1 Jan 2018 → …
|Conference||The 8 th Annual Basic Science International Conference|
|Period||1/01/18 → …|
- Health, Maternal pregnant deaths, Overdispersion, Zero-Inflated Poisson regression