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

BMC Genomics

Table 7 Comparisons with LASSO and DAMBE on simulated datasets: mean NRF values and p-values obtained in the statistical testing of PhyloMissForest samples over the alternative approaches. Lower NRF values denote better quality. N/A denotes scenarios where DAMBE was not able to find any suitable solution

Dataset	%Missing	NRF scores			p-values
		PhyloMissForest	LASSO	DAMBE	vs. LASSO	vs. DAMBE
40x40	5%	2.43	20.54	N/A	0.00	N/A
	10%	7.30	21.89	N/A	0.00	N/A
	15%	7.43	20.00	N/A	0.00	N/A
	20%	12.97	22.70	N/A	0.00	N/A
	25%	12.57	23.24	N/A	0.00	N/A
	30%	13.65	21.35	N/A	0.02	N/A
	35%	20.95	28.38	N/A	0.00	N/A
	40%	17.16	24.59	N/A	0.00	N/A
	45%	19.19	31.89	N/A	0.00	N/A
	50%	16.08	29.46	N/A	0.00	N/A
	55%	27.84	32.97	N/A	0.03	N/A
	60%	25.68	31.89	N/A	0.00	N/A
201x201	10%	14.70	33.84	N/A	0.00	N/A
	15%	18.89	35.56	N/A	0.00	N/A
	20%	21.16	34.85	N/A	0.00	N/A

Bold values in the “NRF scores” columns denote the best NRF scores in the comparison, while in the “p-values” columns they refer to p-values denoting statistically significant improvements

ISSN: 1471-2164