Modelling the number of abortions having excess zero values using zero-inflated generalized Poisson regression
Abdullah Yeşilova, Numan ÇİM, Yıldız Bora
The purpose of this study was to plan for zero-inflated generalized Poisson regression (ZIGP) in the modelling of abortion data that include excess values of zero. The data were collected using the questionnaire technique. It was ascertained that 68.67% (206 observations) of the total number of abortions taken as a model-dependent variable had zero values. A fterward, ZIGP was used to model the dataset. The results of ZIGP(?i, ?, ?i), as mean regression and zero-inflated regression, were determined in two stages. As independent variables were taken into account, it was obtained that zero-inflated data had an important effect on abortion numbers. Therefore, the zero-inflated level was found to have a statistically significant effect (p < 0.01). It was determined that over dispersion did not have an important effect on abortion numbers (p>0.05). In mean regression, the effects of age, number of pregnancies and educational level were found to be statistically significant on abortion numbers (p < 0.01).