Table 3.

Accuracy measures compared between the binary mixture model and the four cut-off choices (mean plus 2 × SD, or 3 × SD, each in both linear and log10 scale) with RT-qPCR. For the binary mixture model classification, the positivity of the rats was determined using the probability of positivity $Pr\left(\mathrm{pos}\right|x)>0.5$. In some situations, the measures are not defined, e.g., in the captive rat study for cut-off values of the mean plus three standard deviations, the positive predictive value is not defined, since all rats were predicted negative. The theoretical binary mixture accuracy measures were not shown for the wild rat population, since they are not relevant for an all-negative population.

	Accuracy	Sensitivity	Specificity	PPV	NPV
Captive study
Binary mixture P(pos) > 0.5	$0.97$	$0.71$	$0.99$	$0.91$	$0.98$
Binary mixture theoretical	$0.98$	$0.93$	$0.99$	$0.81$	$1.00$
Cutoff2	$0.98$	$0.86$	$0.99$	$0.92$	$0.99$
Cutoff2 linear	$0.95$	$0.93$	$0.96$	$0.65$	$0.99$
Cutoff3	-	$0.00$	$1.00$	-	$0.92$
Cutoff3 linear	$0.98$	$0.93$	$0.99$	$0.87$	$0.99$
Feeder study
Binary mixture P(pos) > 0.5	$0.72$	$1.00$	$0.00$	$0.72$	-
Binary mixture theoretical	$0.99$	$1.00$	$0.97$	$0.99$	$0.99$
Cutoff2	$0.72$	$1.00$	$0.00$	$0.72$	-
Cutoff2 linear	$0.72$	$1.00$	$0.00$	$0.72$	-
Cutoff3	-	$1.00$	$0.00$	$0.72$	-
Cutoff3 linear	$0.72$	$1.00$	$0.00$	$0.72$	-
Wild rats
Binary mixture P(pos) > 0.5	$0.71$	-	$0.71$	-	-
Cutoff2	$0.70$	-	$0.70$	-	-
Cutoff2 linear	$0.67$	-	$0.67$	-	-
Cutoff3	-	-	$0.71$	-	-
Cutoff3 linear	$0.70$	-	$0.70$	-	-