. Author manuscript; available in PMC: 2014 Jan 1.

Published in final edited form as: Methods Mol Biol. 2013;972:121–139. doi: 10.1007/978-1-60327-337-4_8

Table 1.

Comparison of parameter estimates of three different procedures for four sets of simulations with different percentages of DE genes (q).

		Percentage of DE genes (q)
Method	Parameter	0.12 (0.005)	0.19 (0.008)	0.36 (0.009)	0.49 (0.008)
HMRF	$\hat{η}$	0.38 (0.026)	0.40 (0.028)	0.41 (0.018)	0.44 (0.020)
	$\hat{ν}$	13.01 (0.061)	13.06 90.061)	13.12 (0.059)	13.18 (0.057)

HMRF-I	$\hat{η}$	0.31 (0.019)	0.34 (0.017)	0.37 (0.013)	0.39 (0.012)
	$\hat{ν}$	13.03 (0.060)	13.06 (0.060)	13.15 (0.057)	13.22 (0.056)

EB	$\hat{η}$	0.067 (0.004)	0.053 (0.003)	0.042 (0.002)	0.039 (0.001)
	$\hat{ν}$	7.27 (0.21)	7.43 (0.21)	7.86 (0.23)	8.21 (0.25)

HMRF: the proposed HMFR model and the ICM algorithm using the network structures; HMRF-I: the proposed HMFR model and the ICM algorithm without using the network structures; EB: the empirical Bayes method of of Tai and Speed (2006). Parameter estimates are averages over 100 simulations; standard error is shown in parentheses. The true parameters are (η, ν)=(0.5,13).