Modeling Determinants of Discriminatory Attitude towards HIV/AIDS Patients in Zambia: A Generalized Additive Mixed Model

Abstract

The 2007 Zambia Demographic Health Survey (ZDHS) HIV/AIDS data was used to investigate the attitude of people to HIV/AIDS patients in Zambia using the generalized additive mixed model. The model was used to simultaneously measure the fixed, nonlinear and random effects. The fixed effects of categorical covariates were modelled using the diffuse prior, P-spline with second-order random walk for the nonlinear effect of continuous variable while the exchangeable normal priors were used for the random effects of the district. The Binomial distribution was used to handle the dichotomous nature of the three dependent variables considered. The dependent variables are stigmatization towards relative sick with HIV/AIDS, teacher who has HIV/AIDS but is not sick and a food vendor who has HIV/AIDS. We found that the people who seem to be responsible for discriminatory attitude towards PLWHA in Zambia are predominantly those who stay in the Southern and Lusaka provinces, the urban areas, who are from at least the middle class wealth index, elite, age 20+ and who either are married/living together with partner.

Keywords: Bayesian inference, binomial, generalized additive mixed model, HIV/AIDS,
Zambia.