Page 245 - Special Topic Session (STS) - Volume 2

STS489 Chibuzor C. N. et al.
(uncorrelated) spatial effects, time main effect as a smooth function,
() and an interaction term, (, ).

Statistical analysis and inference were carried out in Bayesian framework via
Markov Chain Mote Carlo (MCMC) techniques and implemented in R statistical
programming software through its R interface to BayesX known as R2BayesX
[8]. Model fit and complexity were tested using Deviance Information Criteria
(DIC) proposed by [9], the smaller the better.

3. Results
Figure 1 shows the crude (observed) FGM/C prevalence across the survey
years in Kenya, Nigeria and Senegal, indicating a clear picture of geographical
variations in the practice. Red colour indicates highest prevalence regions or
states, while green colour indicates lowest prevalence regions or states.

Figure 1: Evolution of 0-14-year-old girl’s FGM/C prevalence in Kenya, Nigeria and
Senegal

In Table 1, we present the posterior odds ratio (POR) from the fully
adjusted model which accounted for other confounders including temporal,
spatial and spatio-temporal effects. The results shows that in Kenya, girls who
lived in urban region were more likely to be cut than their counterparts.
However, in Nigeria and Senegal, rural girls had higher likelihood of being cut.
The likelihood of FGM/C was not significantly influenced by household wealth
index in Kenya and Senegal, however, girls from lowest wealth quintile
households were more likely to be cut. A girl who professed Muslim faith had
higher likelihood of FGM/C than her Christian counterpart in Kenya. Across the
three countries, we found that a girl’s likelihood of being cut increased if her
mother was circumcised and if her mother her poor level of educational
attainment.
In Figure 2, we show identified and mapped hotspots (red) across the three
countries where the observed FGM/C prevalence were largely due to
unobserved effects of geographical locations of the respondents.

234 | I S I W S C 2 0 1 9