Figure 1.
Monthly malaria case counts and rainfall in Gampaha District over time.
Panel A shows monthly malaria case counts and panel B shows monthly rainfall.
Table 1.
Akaike’s information criterion (AIC) for selected (Gaussian) models on Box-Cox transformed data.
Table 2.
Selection criteria statistics for selected negative binomial models.
Table 3.
Parameter estimates (mean and 95% credible interval) of selected negative binomial models.
Figure 2.
Posterior predictive distributions for the last 12 months of the Gampaha malaria case count series.
In each panel, representing each a month in the last year of the series, the black and the red lines are the outline histogram of the density of the posterior predictive distribution of the negative binomial model and a (Bayesian) Gaussian
model on Box-Cox transformed data, respectively. Models were fitted to the entire data set. In each panel, the observed case count is represented by a blue dot.
Figure 3.
Cumulative distribution function of randomized cumulative probabilities.
The black line represents the cumulative distribution function of randomized cumulative probabilities of the model on monthly numbers of malaria cases in Gampaha, Sri Lanka. The red line represents the cumulative distribution function of randomized residual probabilities of the Gaussian
model on Box-Cox transformed data. The light grey diagonal line (cumulative distribution equals randomized probability) represents on average appropriate predictive distributions. Dotted lines represent 95% confidence boundaries for proportions equalling probability. A: for the last 392 months in the series. B: for the last fifty months in the series.
Figure 4.
Normal Q-Q plot of normalized randomized quantile residuals of the selected model.
Figure 5.
Plot of normalized randomized quantile residuals of the model against the logarithm of relative change.
Monthly malaria case counts were logarithmically transformed after adding one. Then for each month, the difference between this value and the value for the previous month was taken. The diagonal is the fitted regression line.
Figure 6.
Plot of the autocorrelation function of normalized randomized quantile residuals of the selected model.