Bayesian Hierarchical Compositional Models for Analysing Longitudinal Abundance Data from Microbiome Studies

Table 5

Deviance information criterion (DIC) defined in [63] and the DIC defined in [64]. The root mean square deviation (RMSD) of the results obtained fitting the model and the RMSD of the results obtained using the model for prediction.

Data	Model		pD (2002)	DIC (2002)	pD (2004)	DIC (2004)	RMSD (fitting)	RMSD (predicting)
Female families	PM						0.0264	0.0498
	AR(1)-alr						0.0245	0.0398
	AR(1)-clr						0.0245	0.0395
	alr-Bal						0.0346	0.0414
	alr-alr-Bal						0.0315	0.0399
Female genera	PM						0.0174	0.0280
	AR(1)-alr						0.0172	0.0368
	alr-Bal						0.0271	0.0216
	alr-alr-Bal						0.0208	0.0207
Male	PM						0.0111	0.0458
	AR(1)-alr						0.0098	0.0278
	AR(1)-clr						0.0089	0.0289
	alr-Bal						0.0144	0.0309
	alr-alr-Bal						0.0142	0.0306
Simulated data	PM						0.0283	0.0401
	AR(1)-alr						0.0236	0.0398
	AR(1)-clr						0.0239	0.0413
	alr-Bal						0.0339	0.0382
	alr-alr-Bal						0.0329	0.0385
Simulated data	PM						0.0190	0.0310
	AR(1)-alr						0.0144	0.0327
	AR(1)-clr						0.0147	0.0331
	alr-Bal						0.0257	0.0308
	alr-alr-Bal						0.0252	0.0309
Simulated data	PM						0.0001	0.0143
	AR(1)-alr						0.0009	0.0142
	AR(1)-clr						0.0009	0.0124
	alr-Bal						0.0052	0.0060
	alr-alr-Bal						0.0050	0.0061