PhyloMissForest: a random forest framework to construct phylogenetic trees with missing data

BMC Genomics

Table 5 Statistical testing of NRF results for the 40x40 and 201x201 datasets, with regard to autoencoder (AE) and matrix factorization (MF). Statistically significant improvements achieved by PhyloMissForest (under any of the considered configuration profiles) are denoted as \(\checkmark \), while non-significant differences are marked with ×. N/A refers to situations where matrix factorization did not finish execution in an experimental time window of 48 hours

Dataset	%Missing	Non-bootstrap p-values		Bootstrap p-values		PhyloMissForest diff.
		Vs. AE	Vs. MF	Vs. AE	Vs. MF	Vs. AE	Vs. MF
40x40	5%	0.69	0.05	0.69	0.05	×	\(\checkmark \)
	10%	0.31	0.69	0.42	0.42	×	×
	15%	0.05	0.69	0.05	1.00	\(\checkmark \)	×
	20%	0.05	0.31	0.10	0.22	\(\checkmark \)	×
	25%	0.01	0.69	0.01	0.55	\(\checkmark \)	×
	30%	0.10	1.00	0.10	0.31	\(\checkmark \)	×
	35%	0.84	0.15	0.05	0.31	\(\checkmark \)	×
	40%	0.05	0.01	0.03	0.05	\(\checkmark \)	\(\checkmark \)
	45%	0.01	0.05	0.10	0.55	\(\checkmark \)	\(\checkmark \)
	50%	0.01	0.01	0.15	0.22	\(\checkmark \)	\(\checkmark \)
	55%	0.10	0.84	0.84	0.42	\(\checkmark \)	×
	60%	0.01	0.03	0.01	0.84	\(\checkmark \)	\(\checkmark \)
201x201	10%	0.05	N/A	0.05	N/A	\(\checkmark \)	N/A
	15%	0.01	N/A	0.01	N/A	\(\checkmark \)	N/A
	20%	0.01	N/A	0.01	N/A	\(\checkmark \)	N/A

ISSN: 1471-2164