Skip to main content

Table 4 Marginal and conditional coefficients of variation (R2) for OC-COPD models and Lasso-penalized GLMM variants. The two-step LassoGLMM method, in columns 1 and 2, is presented here. The original LassoGLMM, in columns 3 and 4, omits the first step of correlation-based variable screening, adding all OTUs to the LassoGLMM. The GLMM with correlated genera, in columns 5 and 6, uses the correlation-based variable screening step, adding only those variables that are correlated with the outcome to the model, but modifies the second step to not include the lasso penalty. Each method column contains the marginal and conditional R2 that represent fit of the fixed effects and entire model, respectively

From: Measuring associations between the microbiota and repeated measures of continuous clinical variables using a lasso-penalized generalized linear mixed model

  Two-step LassoGLMM Original LassoGLMM GLMM with correlated genera
  Marginal R2 Conditional R2 Marginal R2 Conditional R2 Marginal R2 Conditional R2
BUN (O2) 0.58 0.60 No non zero coefficients All correlated variables were in Two-step LassoGLMM
IGM (O3) 0.19 0.89 No non zero coefficients All correlated variables were in Two-step LassoGLMM
SGOT (O7) 0.22 0.84 No non zero coefficients 0.50 0.59
SGPT (O8) 0.44 0.75 No non zero coefficients All correlated variables were in Two-step LassoGLMM
Cholesterol (O9) 0.80 0.93 0.95 0.98 0.99 1.00