Box plots of error rates of the reconstruction of 100 networks using linear Gaussian DBNs. Each network consists of 30 genes. Error rates are determined by comparing the best reconstructed network with the true network. All three plots are based on the same number of data points per gene. The boxes show lines at the lower quartile, median, and upper quartile values. Outliers are indicated by open circles. (a) One time series replicate with 40 data points per gene based on a random perturbation of the steady state. (b) 10 independent time series replicates each with 4 data point per gene based on random perturbations of the steady state. (c) 10 time series replicates of length 4 each based on single TF knock-outs. FNR: false-negative rate. FPR: false-positive rate. FDR: realized false discovery rate. The observational noise level is 2%.