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

Table 4

Percentage of credible intervals that do not contain zero and the number of parameters in each equation of the model for all databases. S1, S2, and S3 denote the first, second, and third simulations. Note that the AR(1)-clr model does not achieve convergence in the F. genus dataset.

Model	Information	F. families	F. genus	Male	S1	S2	S3
AR (1)-alr	% zeros	12.5%	6.86%	0%	3.80%	1.57%	0%
	Param/equation	9	18	9	15	20	60
alr-alr-Bal	% zeros	41.66%	39.21%	4.16%	11.90%	12.28%	0%
	Param/equation	3	3	3	3	3	3
PM	% zeros	9.54%	8.33%	4.44%	7.69%	0.1%	0%
	Param/equation	6	6	5	13	13	15
AR (1)-clr	% zeros	2.22%		0%	4.16%	4.76%	0%
	Param/equation	10	19	10	16	21	61
alr-Bal	% zeros	43.75%	35.29%	31.25%	14.28%	15.78%	0%
	Param/equation	2	2	2	2	2	2