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

To provide insight into the relative importance of spatial and transport-associated transmission for PED, we applied the correlation analysis to data from the 2013 epizootic. We found that lag-1 cross correlations of positive accessions were positively correlated with the logarithm of transport flows. This relation held whether flows and cross correlations were treated as directional (Fig. 3a) , were averaged over both directions ( Supplementary Fig. S6 ), or were ranked ( Supplementary Figs. S7 and S8). The size of the correlation was comparable to those seen in our simulations for high transmission probabilities across network edges (Supplementary Fig. S3 ). The p values for these correlations were all below 0.05 / 6, which means that they would remain significant after using a Bonferroni method of limiting the probability that any false positives occurred in our tests to 0.05. On the other hand, in no case would the correlations between geographic distance and cross correlations remain significant after such a correction. The correlations between the cross correlation matrix and the shared border matrix were similar to or weaker than those of the distance matrix, were not significant after Bonferroni correction, and were not included in the plots to keep them simpler. Analogous results hold for lag-0 and lag-2 cross correlations. Although higher lags are potentially informative, we did not attempt to analyse them as they were typically smaller, had more skewed distributions, and had greater uncertainty.