Using heterogeneity in the population structure of U.S. swine farms to compare transmission models for porcine epidemic diarrhoea

Eamon B. O’Dea; Harry Snelson; Shweta Bansal

Since the data were highly aggregated and noisy, it was instructive for us to evaluate the performance of our correlation analysis under these conditions. To that end, we simulated the spread of a disease among individual U.S. swine farms, simulated a series of positive accessions, and then performed the correlation analysis on the simulated accession data. The model of disease spread included contact networks based on the spatial and transport structure of U.S. farms. The correlation analysis estimated the strength of association between the similarity of the time series of accessions between a pair of states with the measure of similarity according to either spatial or transport structure. Using previously described methods, 31, 32 we computed global sensitivity indices up to second order for the mean value of these correlations for the parameter space given in Table 1 . These indices estimate the fraction of the variance in the correlation that is due to variation in individual parameters or pairs of parameters. This exercise did not validate our modeling assumptions, but rather evaluated self-consistency by showing the extent to which we could quantify and distinguish between different modes of spread given our assumed models of spread and observation. The procedure is described in further detail in the Supplementary Note.