BMC Bioinformatics

Table 1 Simulation results: estimation performance for \(N=50\) observations, \(\dot{z}_i =\) 10,000 reads, \(T=50\) compositional elements at varying levels of misclassification, and with overdispersion parameter \(d = 0.01\)

From: Analyzing microbiome data with taxonomic misclassification using a zero-inflated Dirichlet-multinomial model

Null
	\(\beta _{\eta _{t1}} \ne 0\)		\(\beta _{\eta _{t1}} = 0\)		\(\beta _{\gamma _{t1}} \ne 0\)		\(\beta _{\gamma _{t1}} = 0\)		\(\pi _t(\varvec{x}_{ip})\)
	ABS	COV	ABS	COV	ABS	COV	ABS	COV	ABS	COV
MicroMiss	0.352	0.945	0.271	0.929	0.115	0.685	0.144	0.761	0.115	0.886
ZIDM	0.350	0.946	0.270	0.934	0.261	0.543	0.209	0.621	0.151	0.565
DM	–	–	–	–	0.796	0.087	0.182	0.909	0.565	0.391
Low
	\(\beta _{\eta _{t1}} \ne 0\)		\(\beta _{\eta _{t1}} = 0\)		\(\beta _{\gamma _{t1}} \ne 0\)		\(\beta _{\gamma _{t1}} = 0\)		\(\pi _t(\varvec{x}_{ip})\)
	ABS	COV	ABS	COV	ABS	COV	ABS	COV	ABS	COV
MicroMiss	0.354	0.941	0.288	0.927	0.273	0.610	0.164	0.851	0.216	0.776
ZIDM	0.720	0.847	0.468	0.956	0.911	0.000	0.124	0.915	0.625	0.216
DM	–	–	–	–	0.797	0.005	0.081	0.984	0.561	0.246
Medium
	\(\beta _{\eta _{t1}} \ne 0\)		\(\beta _{\eta _{t1}} = 0\)		\(\beta _{\gamma _{t1}} \ne 0\)		\(\beta _{\gamma _{t1}} = 0\)		\(\pi _t(\varvec{x}_{ip})\)
	ABS	COV	ABS	COV	ABS	COV	ABS	COV	ABS	COV
MicroMiss	0.359	0.941	0.304	0.938	0.304	0.571	0.157	0.894	0.249	0.728
ZIDM	0.859	0.938	0.291	0.990	0.871	0.012	0.173	0.752	0.603	0.208
DM	–	–	–	–	0.774	0.005	0.076	0.983	0.547	0.234
High
	\(\beta _{\eta _{t1}} \ne 0\)		\(\beta _{\eta _{t1}} = 0\)		\(\beta _{\gamma _{t1}} \ne 0\)		\(\beta _{\gamma _{t1}} = 0\)		\(\pi _t(\varvec{x}_{ip})\)
	ABS	COV	ABS	COV	ABS	COV	ABS	COV	ABS	COV
MicroMiss	0.392	0.942	0.358	0.924	0.344	0.485	0.155	0.931	0.278	0.691
ZIDM	0.974	0.974	0.181	0.998	0.789	0.007	0.126	0.830	0.563	0.211
DM	–	–	–	–	0.758	0.006	0.073	0.984	0.539	0.223

ABS, absolute value of the difference between the estimated and true parameters; COV, 0.95 coverage probabilities

Back to article page

ISSN: 1471-2105

Contact us

General enquiries: journalsubmissions@springernature.com