MOESM7 of Time series analysis of malaria in Afghanistan: using ARIMA models to predict future trends in incidence

Additional file 7: Annex 2. Pairwise correlation between malaria ARIMA model residuals and external regressor residuals at different lags, after pre-whitening (removing trends and seasonality and fitting ARIMA models to each) (first table). In preliminary analyses, statistically significant correlation was observed between rain and humidity (r = 0.7032, p < 0.001); subsequently, humidity was dropped after it was found not to add meaningful information. Had we not performed pre-whitening, statistically significant correlations existed between malaria and other variables at every lag we analyzed.