A Spatial Analysis of Age at Sexual Initiation Among Nigerian Youth as a Tool for HIV Prevention: A Bayesian Approach

Abstract

A study into the geographical variations of sexual initiation among Nigerian youth, age 15–24 years was carried out using a dataset from the 2005 National HIV/AIDS and Reproductive Health Survey in Nigeria. Spatial pattern at highly disaggregated level of state/district of residence as well as non linear effects of observed metrical covariates were explored. Influence of cluster information was explored at two hierarchical levels (census blocks nested within ethnic groups) to assess impact of random effects (frailties). Effects of all categorical covariates were assumed to be linear and hence estimated in a usual parametric form. Inference was based on Bayesian Markov chain Monte Carlo (MCMC) simulation techniques. Appropriate priors were assigned to all effects. Model diagnostic was based on the Deviance Information Criterion (DIC). Time to experience of first sexual intercourse was found to be associated with gender, current age, level of education and rural-urban location of respondents as well as with sexually transmitted infections. Findings also revealed substantial geographical variations on age at sexual initiation. Model that controlled for unobserved heterogeneity due to census blocks was observed to be more adequate than the ones that ignored it, while ethnic group did not seem to provide obvious frailty information in the data. However, model that included census blocks frailty nested within ethnic groups was most superior to all other models in terms of the DIC.