Fig. 5From: Forecasting Chikungunya spread in the Americas via data-driven empirical approachesVariability during Chikungunya forecasting. Percentage failure among countries in the Americas (boxplots) was measured from predictions between August 2014 and January 2015 to assess model predictions developed from six to one month of prior PAHO reports in February 2015, on which predictions were based. The metric identify the match between real cases reported the last month of the study and models developed six or one month in advance (from left to right). Negative values represent under prediction (i.e., cases below the real report) and positive values represent overprediction (i.e., cases above the real report). Failure = 0 represents prediction matching the real number of cases reported. Notice that late models developed with more data accumulated were more close to the real reportsBack to article page