Table 2.

Wall-clock running time (sec) across methods on selected representative simulated conditions on $n=101,201,1001$ for k ranging in 50, 200, 1000. Data points show averages across 50 replicates. ASTRAL did not finish on 24 out of the 50 replicates within four hours for $k=1000$ on MC11H, and thus the data point was omitted. The methods, sorted by the fastest to the slowest, are almost always wASTRID-s, ASTRID, wASTRAL-h, and ASTRAL across all shown conditions. ASTRID and wASTRID-s are much faster than ASTRAL and wASTRAL-h

Running time (s)	k (# genes)	ASTRAL	ASTRID	wASTRAL-h	wASTRID-s
S100 ($n=101$)	50	12.1	0.1	4.6	0.1
	200	29.6	0.2	6.6	0.1
	1000	300.7	1.4	22.8	0.2
MC5 ($n=201$)	50	20.3	0.2	12.3	0.1
	200	57.5	0.6	25.6	0.2
	1000	600.7	1.8	106.9	0.5
MC11H ($n=1001$)	50	552.9	18.0	611.8	11.3
	200	2059.6	14.9	1333.4	8.8
	1000	–	27.5	7191.0	11.8