Comparison of error in chikungunya case count estimates for the heuristic model, baseline nowcasting model and nowcasting model using Google searches.

We compare the mean absolute errors (MAEs) for the chikungunya nowcasts produced by the heuristic model, baseline model and model using Google searches. The first two columns show results across the sample as a whole, reported in number of cases and relative improvement in comparison to the heuristic model. The baseline nowcasting model is far more accurate than the heuristic model, reducing MAE by 34%. The model using Google searches improves further upon the baseline, reducing MAE by 41% relative to the heuristic model. The third and fourth column show results when only considering epidemic periods. Here we see that the error reductions offered by both the baseline model and model using Google searches increase further: the baseline model reduces MAE relative to the heuristic model by 35%, and the model using Google searches reduces MAE by 43%.