TABLE 7.

Comparison of the prior distributions for inapparent CHIKV infections by way of distance and similarity measures/metrics^a

Prior distribution categoryb	Kullback-Leibler divergence (range, 0, +∞)c	Hellinger distance (range, 0, 1)d	Bhattacharyya coefficient (range, 0, 1)e
Overall	0.49	0.41	0.83
Epidemics caused by the Asian lineage	3.34	0.79	0.37
Epidemics caused by the ECSA lineage	0.19	0.25	0.94
Epidemics caused by the IOL strain of the ECSA lineage	0.19	0.24	0.94
Epidemics caused by non-IOL strains of the ECSA lineagef

^aThe Kullback-Leibler divergence, the Hellinger distance, and the Bhattacharyya coefficient quantify differences in global behaviors between two probability distribution functions.

^bFor these analyses, prior distributions were compared to beta(2.97, 16.72), the prior distribution corresponding to all studies that reported a proportion of inapparent CHIKV infections within the expected 3 to 28% range.

^cThe lower the Kullback-Leibler divergence, the more the probability distribution under consideration behaves in a fashion similar to that of the distribution built from studies conforming to the expected range in expectation, and vice versa.

^dThe lower the Hellinger distance (the probabilistic analogue to the Euclidean distance between two points), the more similar the distribution functions, and vice versa.

^eConversely, a low Bhattacharyya coefficient indicates a low degree of overlap between the distribution functions considered, and vice versa.

^fHyperparameters for a prior distribution could not be estimated via maximum likelihood or method-of-moments estimators from the single non-IOL-strain ECSA study identified in the systematic literature search, so no parameters could be calculated.