Table 3.

Estimates of the escape rate in the model where the initial frequency of the escape variant is constrained to be lower than f_c = 10⁻⁴. In the 1^st column we list the epitopes in which escape occurs. In the 2^nd column we list estimates of the escape rate assuming a constant escape rate and using eqn. (10) and eqn. (16) with a penalizing term 10^{105(f₀−f_c)} added to the log likelihood. Model fits are shown in Figure 3A. In the 4^th and 5^th columns we list estimates of the initial escape rate and the rate of decline of the escape rate assuming that the escape rate declines over time using eqn. (13) and eqn. (16). Fits are shown in Figure 3B.

Model	Constant ε	Decreasing ε
epitope	ε, day−1	ε0 day−1	a, day−1
Pol80*	0.06	2.55	0.09
Rev49	0.05	0.07	0.01
Vif113*	0.03	0.76	0.02
Gag389	0.27	0.53	0.04
Nef185*	0.51	1.04	0.08

For the 3 epitopes indicated by * allowing the escape rate to change over time significantly improved the quality of the model fit to data (likelihood ratio test, p < 0.0001).