The same techniques described above for spline correlogram estimation were used to obtain smoothed, non-parametric estimates of R T t 0 (d), confidence intervals for experimental estimates, and regions of significance (see Text S1 for more details).