Table 3. Power of the linear regression and GLM-based test statistics for different co-expression dynamics and sample sizes.

co-expression dynamics	sample size (n)	linear regression ⋆	GLM-based LR test ⋆	GLM-based Larntz & Perlman ⋆
linear positive association	125	0.408 [0.377, 0.439]	0.314 [0.285, 0.343]	0.188 [0.163, 0.213]
linear positive association	450	0.635 [0.605, 0.665]	0.826 [0.802, 0.850]	0.797 [0.772, 0.822]
linear positive association	800	0.712 [0.683, 0.741]	0.990 [0.983, 0.997]	0.989 [0.982, 0.996]
linear negative association	125	0.451 [0.420, 0.482]	0.300 [0.271, 0.329]	0.184 [0.159, 0.209]
linear negative association	450	0.621 [0.590, 0.652]	0.838 [0.815, 0.861]	0.819 [0.795, 0.843]
linear negative association	800	0.723 [0.695, 0.751]	0.988 [0.981, 0.995]	0.987 [0.979, 0.995]
non-linear association (parabola)	125	0.219 [0.193, 0.245]	0.293 [0.264, 0.322]	0.243 [0.216, 0.270]
non-linear association (parabola)	450	0.051 [0.037, 0.065]	0.759 [0.732, 0.786]	0.856 [0.834, 0.878]
non-linear association (parabola)	800	0.010 [0.003, 0.017]	0.969 [0.958, 0.980]	0.993 [0.987, 0.999]
non-linear association (wave)	125	0.253 [0.226, 0.280]	0.348 [0.318, 0.378]	0.193 [0.168, 0.218]
non-linear association (wave)	450	0.152 [0.129, 0.175]	0.863 [0.841, 0.885]	0.841 [0.818, 0.864]
non-linear association (wave)	800	0.108 [0.088, 0.128]	0.992 [0.986, 0.998]	0.990 [0.983, 0.997]
weak positive association	125	0.278 [0.250, 0.306]	0.143 [0.121, 0.165]	0.072 [0.055, 0.089]
weak positive association	450	0.257 [0.229, 0.285]	0.182 [0.158, 0.206]	0.183 [0.159, 0.207]
weak positive association	800	0.235 [0.208, 0.262]	0.316 [0.287, 0.345]	0.351 [0.321, 0.381]

Data simulated for a four-gene module.

^⋆ estimate [95% confidence interval]