Page 96 - Contributed Paper Session (CPS) - Volume 3

CPS1952 Michele N. et al.
ADMIN2 units in Colombia, 21 ADMIN1 units in Venezuela, 1 ADMIN1 units in
Panama and the ADMIN0 (national) level for Ecuador, Suriname and Paraguay.
The model chosen by backwards regression was refitted to all of the data since
both training and test coverage probabilities (90.244% and 90.204%) are close
to the target of 95% and the root mean squared errors (0.0211 and 0.0187)
are comparably low.
Table 1 shows the parameter estimates for the chosen monthly proportion
model. Under a 5% significance level, we observe a positive relation between
monthly proportions and EVI but negative relations with CHIRPS, CHIRPS_lag1
and EVI_lag2. The negative relation with rainfall could be explained by the
increase in mosquito breeding habits when river recede during the dry season
(Valle & Lima 2014) Intense rain could also reduce treatment seeking rates and
hence the number of recorded malaria cases.
Figure 1 shows the estimates of the seasonality index which were
computed using 2016 Pv API estimates and 100 realisations of the estimated
monthly proportions. The strongest seasonality based on the mean index is
observed in Colombia, near its borders with Brazil and Peru. It was also found
that most of the LAC region experiences only one transmission season per
year. As an example, we present the map of the mean start month of the first
season and its standard deviation in Figure 2. These are calculated via the
circular definitions and transformed back into months via multiplication with
12 (Pewsey et al. 2013). The mean start months are reasonably contiguous
2
across space as expected. Such maps of seasonality characteristics will be
useful for policymakers for planning interventions.

Term Median 95% CI Term Median 95% CI
Intercept -2.205 (-2.276 , -2.133) LST_delta_lag1 -0.017 (-0.116, 0.082)
CHIRPS_lag3 0.014 (-0.010, 0.038) TCB_lag1 -0.026 (-0.085, 0.033)

CHIRPS -0.065 (-0.093, -0.036) EVI 0.092 (0.019, 0.165)
TCB_lag3 0.014 (-0.042, 0.071) EVI_lag2 -0.124 (-0.203, -0.044)
TSI_Pv_lag3 -0.070 (-0.157, 0.017) obs. var ( 2 ) 0.188 (0.173, 0.203)
2
LST_night 0.035 (-0.080, 0.149) spde.var.nom ( ) 1.167 (1.057, 1.288)

CHIRPS_lag1 -0.041 (-0.070, -0.012) spde.range.nom () 5.350 (4.954, 5.750)
LST_delta_lag3 0.094 (-0.005, 0.193) AR.rho () 0.908 (0.892, 0.922)
Table 1: Posterior medians and 95% credible intervals (CIs) for the
parameters of the refitted model.

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