DESIGN-BASED APPROACH TO MODELLING OF MALARIA INCIDENCE IN NIGERIA.

Abstract

Most available datasets on malaria in Nigeria used in literatures, were limited to small localised
population or small sample size of cohorts. Hence, inference were not scalable to the entire

country. Interest in malaria modelling is on population estimates and in order to improve cost-
effectiveness and representational accuracy in large-scale surveys, complex sampling designs such

as multistage sampling with uneven selection probabilities are often employed. To avoid biased
estimations, specific, weighted analytical techniques are applied to take into consideration
complex data structures. Such techniques, like resampling or Taylor series linearization, give
estimates about the population, instead of the sample as seen in model-based analysis. The
literature on modelling malaria in Nigeria used survey datasets that were collected from complex
survey designs, yet scarce literature accounted for the complex design in their analyses. The
present study examined the effects of complex design in modelling malaria incidence in Nigeria,
using dataset from the 2021 Nigeria Malaria Indicator Survey (NMIS). The effects of sampling
weights and survey information in the analysis of malaria datasets were examined and compared
with model-based approach. The analysis capturing the survey design produced a different result
from the model-based approach, effectively showing the difference in the sample and population
estimates, for which the latter is the object of surveys such as the 2021 NMIS. Hence, this paper
adds to the literature on country-wise malaria incidence analysis and how survey design is
incorporated in the analysis to meet the objective of such survey.
Keywords: complex design; design-based; malaria incidence; model-based; weights