Right-Censored Poisson Regression Model for Fertility Count Data
Main Article Content
Abstract
Introduction: Count data represents the number of occurrences of an event within a fixed period [1, 2, 3, 4]. For example the number of caesarean-section deliveries in the lifetime of women. Count data is encountered in almost all research areas including economics, medicine, management, industrial organizations, and many more [5]. Count data is very common in various fields such as biomedical science, public health, and marketing. Poisson models are widely used in the regression analysis of count data and as a basis for count data analysis [6, 7, 8, 9, 10,11].
Objectives: The main aim of this study is to estimate the parameters of interest and compare the number of caesarean-section deliveries (NCSD) among women aged 15-49, in the state of Andhra Pradesh, India, using the right-censored Poisson regression model (RCPRM) and right-censored negative binomial regression model (RCNBRM). The fertility count data set, the real-world data of National Family Health Survey (NFHS-5), 2019-2021, from the Demographic and Health Surveys (DHS), 2019-2021 phase VII data is used for the analysis..
Methods: Investigating the delivery patterns among pregnant women. This study develops an algorithm based on Integrated Nested Laplace Approximation (INLA) for fitting the model NCSD in RCPRM and RCNBRM. The response variable NCSD is right-censored at 1, one caesarean-section delivery. The analysis is carried out using the INLA package in R.
Results: By use of the Deviance Information Criterion (DIC) and Watanabe-Akaike information criterion (WAIC), the result shows that the RCPRM; DIC (4467.14) and WAIC (4463.86) present a comparatively better fit in modelling the right-censored NCSD than the RCNBRM; DIC (4468.83) and WAIC (4465.08).
Conclusions: The INLA provides an efficient algorithm to model in RCPRM and RCNBRM. For further research, comparing the RCPRM with other models that estimate over-dispersion in count data is recommended.