Distribution of Genotype network properties in a set of coalescent simulations, varying the number of samples used to construct the networks.

<p>We first simulated 15,000 haploid sequences (5,000 for each of an African, Asian and European population). Then, we randomly sampled a given number of haplotypes, from 5,000 to 100, and calculated the distribution of properties of genotype networks built using only the randomly sampled individuals. Networks are based on 11 SNVs, and each data point represents the average of five random subsamplings of the same size. Bars are too short to be visible for most data points, and show 6 times the standard deviation of the mean.