Skip to main content
Fig. 1 | BioData Mining

Fig. 1

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

Fig. 1

Overview of the two-step LassoGLMM model developed. Species (or genera, OTUs, or any other explanatory variables of interest) are divided into those that are correlated with the dependent continuous variable, Y, and those that are not. Species that are correlated are stored in a matrix X. Relevant categorical variables, found through a review of expert literature or other means, are stored in a matrix W. Indicators of repeated measures such as patient ID are stored in matrix Z. Matrices X, W, and Z are entered into a generalized linear mixed model to be regressed on outcome variable Y. Coefficient β for matrix X and coefficient B for W are subjected to the lasso penalty. Any species that retain non-zero coefficients are considered strongly associated with the dependent variable Y

Back to article page

BioData Mining

ISSN: 1756-0381