biomvRhsmm: Genomic Segmentation with Hidden Semi-Markov Model

Table 4

Processing time and error estimate of the compared models.

	avg.𝑡	Simulation 1			Simulation 2
		avg.cp	MAE	RMSE	avg.cp	MAE	RMSE
hsmm	0.25645	12/11.16	5.756	3.476	6/6.50	18.73	6.31
bcp	1.46298	NA	NA	NA	NA	NA	NA
bioHMM	6.96811	14/14.71	8.655	7.376	NA	NA	NA
CBS	0.12168	12/11.02	7.444	4.178	5/4.93	20.762	7.068
CGHseg	0.28938	12/9.89	9.059	4.783	7/7.00	14.83	5.533
GLAD	0.23725	10/8.22	13.128	6.071	3/4.15	22.139	7.668
HaarSeg	0.00268	17/15.57	12.896	4.984	10/10.65	10.117	4.018
HMM	0.28008	7/38.97	94.666	80.792	1/60.05	144.83	97.178

avg.𝑡 is calculated as the mean run time of 20000 simulation iterations.
avg.cp is the median/mean number of segments estimated across 3 SNR settings.
MAE is calculated as the mean absolute error ∑|no.seg−true.no.seg|/𝑛.
RMSE is the rooted mean squared error ∑(no.seg−true.no.seg)²/𝑛.
NA indicates that the measurement is not applicable for this algorithm. For bcp, the model output posterior means for each position that does not tend to form segments with constant mean. For bioHMM, the model cannot be run, thus no results were collected.