Table 1.

Parameter estimates for the baseline model equation (2.1), assuming a constant recrudescence rate r. ΔAIC is computed relative to the best stylized distribution, the log-logistic, see electronic supplementary material, table S1. Note that when estimating the recrudescence rate from all the data, we used the delay τ estimated using data restricted to rebounds less than 60 days from ATI.

time window	parameters	estimates (95% CI)	ΔAIC
rebound within 60 days of ATI	r = (1 − q)aL0	0.0902 (0.0771,0.1054) per day	n.a. (restricted data)
	delay τ	4.93 (4.24,5.70) days
rebound across all observed times	r = (1 − q)aL0	0.0278 (0.0231,0.0335) per day	252.6 (with one parameter)
	delay τ	4.93 days fixed