Page 51 - IJPS-11-2

International Journal of
Population Studies Single mother spatial distribution in Nigeria

variations across Nigeria and highlight places where single 2.2. Statistical analysis
mothers are more likely to be found; such analysis can We considered a spatiotemporal model within a Bayesian
also aid protective interventions designed specifically framework that allows for the spatiotemporal analysis of
for these individuals. This is particularly important in the pattern of single motherhood in Nigeria. We assumed
a highly heterogeneous country like Nigeria. However, a binomial distribution for the random variable Y that
ij
a countrywide analysis often masks relevant local-level denotes the number of single mothers in state i during the
information (Alaba et al., 2015; Somo-Aina & Gayawan, j survey period, and n is the corresponding number of
th
ij
2019). Consequently, this study aimed to investigate the women who participated in the survey. That is, Y ~ binomial
ij
spatiotemporal changes in single motherhood in Nigeria (n , p ) for i= 1,…,37 and j=1,2,3,4; p is the probability
ij
ij
ij
based on data of the second administrative strata of the (prevalence) of single motherhood. We used the logit of
country. We extracted data from the Nigeria Demographic the probability of the single motherhood outcome given
and Health Survey conducted in four waves and used a as follows:
Bayesian generalized additive spatial model (Blangiardo &
Cameletti, 2015) in which the statistical inference was made logit(p ) = β + vβ + γ + ϑ + δ ij (I)
i
0
ij
j
through a probabilistic approach for Bayesian inference.
The findings from the study elucidate the dynamics of Where β is the model intercept, v is a vector that
0
family formation relating to single motherhood across collects all the categorical variables included as linear
Nigerian states during the period under consideration. terms in the model, β is a vector of the linear parameter,
and γ is the random effect for the state assumed to have a
i
2. Data and methods spatial structure and modeled using a Gaussian intrinsic
2.1. Data source conditional autoregressive prior with the spatially
correlated random term for each state. This was determined
We analyzed data on single motherhood from four waves as the sum of the values assigned as weight based on the
of the Nigeria Demographic and Health Survey conducted neighborhood structure of the spatial units, formulated
in 2003, 2008, 2013, and 2018. The National Population through an adjacency matrix that considers two states as
Commission of Nigeria implemented these surveys with neighbors if they share a common border. The term ϑ is
i
financial and technical assistance from The Demographic a non-linear smooth function assumed for the metrical
and Health Surveys Program. A two-stage sampling covariate age of the women and was modeled assuming
design was used in each wave to determine participating a second-order random walk prior, δ is a spatiotemporal
ij
households for the survey. In the first stage, clusters random effect term, and the parameter vector δ is
(enumeration areas) were chosen from the census data of the considered to have a Gaussian distribution with a precision
country, and in the second stage, participating households matrix τ R , where τ is a structured matrix that indicates
δ
δ
δ
were determined. Women aged 15 – 49 years found in the type of relationship existing between the space and time
the sampled households were eligible to participate in the components. More details about the model are provided
survey, which entailed a detailed questionnaire regarding in a previous study (Blangiardo & Cameletti, 2015). We
the demographic and health conditions of the women and used the deterministic approach to Bayesian inference and
their households. The response variable created for this employed the integrated nested Laplace approximation
analysis is a binary indicator that takes the value of 1 if the available through the R-INLA package (Rue et al., 2009).
woman has never been married or was formerly married/ Other estimation procedures such as using a Markov chain
living with a man and 0 if the woman is currently in a union Monte Carlo simulation are also possible.
or living with a man. The data set was restricted to women
with at least one child, thus excluding single women who 3. Results
were not single mothers. We considered the following Figures 2 and 3 present the maps of the estimated prevalence
variables in our analysis: nature of the place of usual of single motherhood in Nigeria for each of the four survey
residence; the highest level of education; religion; ethnicity; waves. Figure 2 presents the results when no covariates are
household wealth index; whether the woman has access to included in the model, whereas Figure 3 presents results
newspapers, radio, and television at least once a week; and that include all available covariates. The results show that
employment status of the woman. The state in which the in Nigeria, the distribution of single motherhood exhibits
woman resides at the time of the survey was considered as distinct spatial patterns, with significant regional variations
the spatial unit in the model. Administratively, Nigeria is during the years. Initially, in 2003, states in the South-South
divided into 36 states and a federal capital territory, Abuja, and South-East geopolitical zone, such as Rivers, Cross
as reflected in Figure 1. River, Akwa Ibom, Ebinyi, Enugu, and Bayelsa, showed

Volume 11 Issue 2 (2025) 45 https://doi.org/10.36922/ijps.4313