. 2017 Nov 20;13(11):e1006702. doi: 10.1371/journal.ppat.1006702

Table 2. Performance of the estimators of the harmonic mean of effective population sizes $N_{e}^{h} (t_{d})$ and variant fitness r obtained with the two numerical experiments.

Experiment^a	Parameter^b	R²	Intercept	Slope	Accuracy of 90% CI	Mean bias [95% CI]
Experiment 1	${\hat{r}}_{i}$ (all variants)	0.93	0.02	0.98	91%	10⁻⁴ [-0.05;0.05]
Experiment 1	${\hat{N}}_{e}^{h}$ (all dates)	0.86	28	0.91	90%	0.18 [-0.42; 1.48]
Experiment 2	${\hat{r}}_{i}$ (all variants)	0.92	0.06	0.94	89%	8.10⁻⁴ [-0.07;0.07]
Experiment 2	${\hat{N}}_{e}^{h}$ (all dates)	0.85	20	0.87	87%	0.03 [-0.53; 1.17]

^a Experiment 1: 750 simulated datasets with 5 variants under a wide range of selection and genetic drift regimes. Experiment 2: 350 simulated datasets with an additional and undetected sixth variant.

^b For each parameter, the determination coefficient R², the slope and the intercept of the best linear model fit between predicted and true values are given, together with the percentage of true parameter values included in the 90% confidence interval and the mean relative bias and its 95% confidence interval.

Table 2. Performance of the estimators of the harmonic mean of effective population sizes Neh(td) and variant fitness r obtained with the two numerical experiments.

Table 2. Performance of the estimators of the harmonic mean of effective population sizes $N_{e}^{h} (t_{d})$ and variant fitness r obtained with the two numerical experiments.