reality.check: Check simulations for realism

In zellerlab/SIMBA: Simulation of Microbiome Data with Biological Accuracy

Check simulations for realism

Description

This function will compare simulated data with the real data on which the simulations are based

Usage

reality.check(sim.location, group, ml=FALSE)

Arguments

sim.location	file name for the .h5 file containing the simulations
group	character, name of the group within the .h5 file containing the simulated features OR "original" to get the same measures for the original data set
ml	Boolean, indicating if a machine-learning model should be trained to distinguish real and simulated data (see Details below). Defaults to FALSE

Details

This function computes a list containing different measures for how realistic the simulations are, containing

• sparsity measures for the sparsity of samples, computed as L0-norm and as Gini-coefficient

• variance measures for the variance of features, containing additionally measures for differences between groups, such as generalized fold change and prevalence difference

• AUC-BC AUROC value for how well real and simulated data can be distinguished based on the Bray-Curtis distance

• AUC-E AUROC value for how well real and simulated data can be distinguished based on the log-Euclidean distance

• AUC-ML AUROC value for how well real and simulated data can be distinguished based on a SIAMCAT machine learning model

Value

Returns a list of measures

zellerlab/SIMBA documentation built on June 2, 2025, 6:25 a.m.