Fig. 4
From: Improving network inference algorithms using resampling methods
Realistically sized dataset. Number of conditions reduced to 40, and a sweep across subsampling fractions is performed to show performance of BCLR on realistically sized data sets derived from the greater E. coli dataset. Bands indicate a 99% confidence interval constructed from the samples taken at each data fraction. Differences were significant (p-value < 0.01) for subsampling fractions 30% and below