Table 3.

Responses of GC1 and GC2 to changes in GC3, by domain

	n		R2	Slope ± SE	Y at GC3 = 0 ± SE	Y at GC3 = 100 ± SE
Bacteria	311	GC1	0.91	0.370 ± 0.007	0.367 ± 0.004	0.737 ± 0.003
		GC2	0.80	0.219 ± 0.006	0.291 ± 0.004	0.510 ± 0.003
Archaea	28	GC1	0.85	0.38 ± 0.03	0.35 ± 0.02	0.73 ± 0.02
		GC2	0.60	0.16 ± 0.03	0.30 ± 0.01	0.45 ± 0.01
Eukaryotes	257	GC1	0.57	0.24 ± 0.01	0.402 ± 0.008	0.643 ± 0.007
		GC2	0.38	0.15 ± 0.01	0.334 ± 0.007	0.482 ± 0.006

Because there is error in both axes, but there should be a definite causal relationship between GC3 and GC1 or GC2, we use model I regression to predict specific values of GC1 or GC2 from a set value of GC3, and thus to calculate the most likely proportion and GC content of constant sites [73].