Bayesian Semiparametric Modelling of Contraceptive Behaviour in India Via Sequential Logistic Regressions

Abstract

Summary Family planning has been characterized by highly different strategic programmes in India, including method‐specific contraceptive targets, coercive sterilization and more recent target‐free approaches. These major changes in family planning policies over time have motivated considerable interest towards assessing the effectiveness of the different planning programmes. Current studies mainly focus on the factors driving the choice among specific subsets of contraceptives, such as a preference for alternative methods other than sterilization. Although this restricted focus produces key insights, it fails to provide a global overview of the different policies, and of the determinants underlying the choices from the entire range of contraceptive methods. Motivated by this consideration, we propose a Bayesian semiparametric model relying on a reparameterization of the multinomial probability mass function via a set of conditional Bernoulli choices. This binary decision tree is defined to be consistent with the current family planning policies in India, and coherent with a reasonable process characterizing the choice between increasingly nested subsets of contraceptive methods. The model allows a subset of covariates to enter the predictor via Bayesian penalized splines and exploits mixture models to represent uncertainty in the distribution of the state‐specific random effects flexibly. This combination of flexible and careful reparameterizations allows a broader and interpretable overview of the policies and contraceptive preferences in India.

Impact of Sociocultural Factors on Contraceptive Use: A Case Study of Pakistan

“…Thus, to better articulate a more applicable form of the data, we explored the classical logistic regression model [ 24 ] and multinomial logistic regression model [ 6 ] on it. We also provided Bayesian logistic and multinomial regression models [ 25 ] to compare estimates.…”

“…A p value should not be interpreted as the likelihood that the null hypothesis is true, but instead refers to the probability that the data will be observed or even more extreme than when the null hypothesis is true. The likelihood of the values of parameters can be directly obtained in the Bayesian inference by finding, at the right of the region of that value, the area of the posterior distribution, which is equal to the proportion of the values of the parameter in the posterior sample larger than that value [ 25 ]. We may use this data to file the results of Bayesian statistical analysis as a means of estimating parameters with so-called 95% Bayesian credible intervals.…”

“…In the Bayesian viewpoint [ 25 ], we formulate linear regression using probability distributions rather than point estimates. In this paper, we use the Bayesian regression model to predict the outcome of the nonsampled set [ 26 , 27 ].…”

Finite-dimensional Discrete Random Structures and Bayesian Clustering

“…In addition, the Dirichlet multinomial process has been employed in a variety of statistical applications, e.g. for Bayesian modeling of network data (Durante et al, 2017;Durante and Dunson, 2018), for semi-parametric random effects in regression models (Rigon et al, 2019), and for functional data analysis (Dunson et al, 2008;Petrone et al, 2009). Finally, the symmetric Dirichlet for (π 1 , .…”