Use of posterior predictive assessments to evaluate model fit in multilevel logistic regression | Veterinary Research, a journal on Animal Infection

Open Access

Table I.

Sensitivity and specificity of MCMC P values for each prediction method (full = full posterior predictive method, mix 1 and mix 2 = mixed predictive methods 1 and 2 respectively) compared to MCMC P values for cross validation, at different P value thresholds (as specified).

		Cross validation		Total	Sens (%)	Spec (%)
		0	1	Total	Sens (%)	Spec (%)
P value > 0.90 or < 0.10
full	0	86	14	100	17.6	100.0
	1	0	3	3
	Total	86	17	103
mix 1	0	84	3	87	82.4	97.7
	1	2	14	16
	Total	86	17	103
mix 2	0	86	10	96	41.2	100.0
	1	0	7	7
	Total	86	17	103
P value > 0.95 or < 0.05
full	0	93	8	101	20.0	100.0
	1	0	2	2
	Total	93	10	103
mix 1	0	90	5	95	50.0	96.8
	1	3	5	8
	Total	93	10	103
mix 2	0	93	7	100	30.0	100.0
	1	0	3	3
	Total	93	10	103
P value > 0.975 or < 0. 025
full	0	98	5	103	0.0	100.0
	1	0	0	0
	Total	98	5	103
mix 1	0	98	2	100	60.0	100.0
	1	0	3	3
	Total	98	5	100
mix 2	0	98	4	102	20.0	100.0
	1	0	1	1
	Total	98	5	103