Fig. 2From: An expectation-maximization algorithm enables accurate ecological modeling using longitudinal microbiome sequencing dataRobustness of parameter estimation with BEEM. a Results with an increasing number of species but fixed number of replicates (60). As expected, parameter estimation gets harder, but BEEM’s performance tracks the ideal case using BLASSO with noise-free biomass values, especially for interaction parameters. b Median relative error in biomass estimates remains less than 10%. c Results with an increasing number of replicates and fixed number of species (10). BEEM’s performance converges to that of BLASSO with noise-free biomass as the number of replicates increases. d Median relative error in biomass estimates reduces noticeably as the number of replicates increasesBack to article page