Idea 1: We should compare the Mbiome+demo model to a Mbial-Like-Noise+demo model in order to really figure out if microbiome is adding to the demographics or not (and potentially, how much of the mean improtance survives the average importance of noise given each bug's intrinsic scale)
Idea 2: the contrapositive: if we shuffle only the demograhics, will its numerical properties keep bumping up its importance even though we completely destroy the correlation?
Idea 1: We should compare the Mbiome+demo model to a Mbial-Like-Noise+demo model in order to really figure out if microbiome is adding to the demographics or not (and potentially, how much of the mean improtance survives the average importance of noise given each bug's intrinsic scale)