Remotely-sensed, nocturnal, dew point correlates with malaria transmission in Southern Province, Zambia: a time-series study

Table 5 Rural health centre ARIMAX-models’ variables, parameters and forecasts

Model and parameters (p,d,q)	AIC-fit	MAE-test
Maamba (5,0,5)	154.53	33,5%
Siamuleya (2,1,3)	125.57	0,7%
Siatwinda (4,0,2)	166.26	20,7%
Siavonga (1,0,0)	161.20	14,2%
Sinafala (3,0,2)	83.10	1,54%
Maamba variables	Coefficient	p-value
L7 NDVI	0.025	p = 0.982
Siamuleya variables	Coefficient	p-value
L5 LST	0.091	p < 0.05
L7 LST	0.15	p < 0.01
Siatwinda variables	Coefficient	p-value
L6 DWP	0.061	p < 0.001
Siavonga variables	Coefficient	p-value
L7 NDVI	8.09	p < 0.001
Sinafala variables	Coefficient	p-value
L6NDVI	1.13	P = 0.328
L4NDVI	3.38	p < 0.01
Forecast model w37-41	Forecast MAE	Predicted vs actual incidence
Maamba	-37.64%	41.6 vs. 75.1
Siamuleya*	-13.13%	17.1 vs. 22.3
Siatwinda	2.34%	182.6 vs. 191.9
Siavonga	-12.11%	21.8 vs. 29.1
Sinafala	No data	No data

Five rural health centres (RHC) from the high zone were modelled. Akaike’s information criteria (AIC) and mean average error (MAE) are shown for each of the five best-fit ARIMAX models. AIC was calculated for the fitting period. MAE was calculated for the testing period, 2011 w19 through 2012 w18. Normalized differentiated vegetation index (NDVI), nocturnal land-surface temperature (LST) and nocturnal dew point (DWP) are shown with specific lag-times with its coefficients as well as p-values for each RHC. Four-week forecasts MAE and predicted cases vs. actual cases for the four weekly forecasts of the health centre models are shown. Sinafala had insufficient data for the forecast period to be evaluated.
*Siamuleya forecast only three weeks ahead, due to missing data thereafter.

ISSN: 1475-2875