Models of the natural logarithm of nuclear gene copies detected through qPCR as a function of the sample type (Nobservations = 90, Nindividual bats = 12, buccal swab brush or wing punch.

This includes individual WCL025, not shown in S1A Fig. because it only had a wing punch), or the preservation medium for wing punches (Nobservations = 228, Nindividual bats = 21, silica beads, ethanol, or DMSO. This includes individual ALR074, not shown in S1B Fig. because it only had a wing punch preserved in ethanol). No-predictor models serve as null hypotheses of no effect of the sample type or preservation medium on the number of rag2 copies detected using qPCR. Group-level intercepts of qPCR run and/or individual bat sampled serve to capture effects of unmeasured variability from subtle differences in qPCR starting conditions and bat behavior. Significance was measured using a likelihood ratio test of the model with predictors to its null. Models with different sample types or preservation media as predictors can be compared to one another using the Akaike Information Criterion (AIC). The model with the lowest AIC is the best fit to the data. DF = model degrees of freedom, logLik = log-likelihood of model fit to the data, and LRT = likelihood ratio test statistic.Models of the natural logarithm of nuclear gene copies detected through qPCR as a function of the sample type (Nobservations = 90, Nindividual bats = 12, buccal swab brush or wing punch.