TABLE 3.

Hypothesis testing using Bootstrap to compare compositional lasso (CL) with random forests (RF) based on Stability or prediction metric (MSE/AUC) using two experimental microbiome data sets with continuous or binary outcomes (*indicate statistically significant)

Dataset (Continuous)	Estimated mean difference (CL‐RF) in Stability index with 95% CI	Estimated mean difference (CL‐RF) in MSE with 95% CI
BMI Gut	0.27 (0.17, 0.34)*	11.8 (‐2.1, 41.2)
pH Soil	0.36 (0.28, 0.44)*	0.08 (‐0.28, 0.95)

Dataset (Binary)	Estimated mean difference (CL‐RF) in Stability index with 95% CI	Estimated mean difference (CL‐RF) in AUC with 95% CI
BMI Gut	0.30 (0.11, 0.42)*	‐0.09 (‐0.19, ‐0.02)*
pH Soil	0.43 (0.37, 0.5)*	‐0.02 (‐0.08, 0)