Estimating Sex Disparities in Post-natal Survival using Bayesian Methods
Conference
64th ISI World Statistics Congress - Ottawa, Canada
Format: IPS Abstract
Keywords: bayesian hierarchical model, inla, model, model building, sex-specific mortality, time series
Session: IPS 169 - Statistical Research by Women from Around the Globe
Monday 17 July 2 p.m. - 3:40 p.m. (Canada/Eastern)
Abstract
Sex differences in survival have been documented below age 5 but little attention has been devoted to older children and youth aged 5-24. Data on sex differences between ages 5 years and 24 years are sparse in low- and middle-income countries, due to the incompleteness of death registration. Bayesian modeling is useful to estimate sex ratios of mortality in countries with low data availability/quality. We used a Bayesian hierarchical time series model to estimate levels and trends in the sex ratio of mortality for all countries from 1990 to 2021 in the population aged 5-24. The sex ratio of mortality is modeled as the product of the expected sex ratio and country-specific deviations. We use a 2nd order random walk model to estimate the global sex ratio and a 1st order random walk model for temporal fluctuation within each country. The statistical computing is carried out by the Integrated Nested Laplace Approximations (INLA). We demonstrate the model-building process and motivate the choices of functions for the model elements.