Fig. 2
From: MetaNN: accurate classification of host phenotypes from metagenomic data using neural networks
Synthetic microbial frequency count distribution generated using NB distribution based on microbiome profiles. a The underlying true distribution which is highly zero inflated (i.e., no presence of certain microbe). b Type 1 error that adds non-zero noise to the zero count entries in order to change the distribution. c Type 2 error that changes the underlying non-zero entries to zeros. d Type 3 error changes the distribution of non-zeros counts. Note that all different types of errors are added with probability of 0.5