
Analysis of Antenatal Care Visit Data in Bangladesh Using Zero Modified Count Regression Model
Author(s) -
Nasiba Maruf Ahmed,
Taslim Sazzad Mallick
Publication year - 2019
Publication title -
the dhaka university journal of science
Language(s) - English
Resource type - Journals
eISSN - 2408-8528
pISSN - 1022-2502
DOI - 10.3329/dujs.v67i2.54583
Subject(s) - overdispersion , count data , negative binomial distribution , poisson regression , residence , statistics , zero inflated model , regression analysis , mathematics , birth order , medicine , poisson distribution , demography , econometrics , population , environmental health , sociology
In medical science, pharmaceutical studies, public health and socio-economic researches we often encounter the situation of excess of zeros in count data. This preponderance of zeros leads to overdispersion. In such cases traditional count data regression models like Poisson and negative binomial (NB) regression may not be pertinent for inference. The two most commonly used types of model that have been developed to adjust for excessivezeros in count data are Hurdle and zero-inflated models. In this study we have analyzed the antenatal care (ANC) visit data of pregnant women in Bangladesh using traditional and zero-modified count models. Based on the model selection criteria, we found that negative binomial hurdle model fits the data best. Through this analysis,we have perceived that the variables age of mother, division, birth order (order a child is born), place of residence, economic condition, media exposure of the mother, mainaccess road to village and education gap between husband and wife have significant impact on the mean number of ANC visits taken.
Dhaka Univ. J. Sci. 67(2): 117-122, 2019 (July)