I work in the field. The statistical mistakes are only one problem- researchers in the field also make huge mistakes in understanding both microbial diversity and computational genomics. It is partly because many of these studies are run by human biology-minded researchers who know very little about either statistical analysis or microbial diversity (IMO ignorance of microbiology is as big an issue in the field as ignorance of stats).
The other issue is this is a uniquely and enormously complicated intersection of science- there simply are few to no people who are simultaneously knowledgeable about immunology, statistics, computational genomics, microbial physiology, and microbial diversity- the bare minimum number of fields that will come into play when dissecting a host-microbe interaction.
That being said there are some very brilliant and rigorous people working in the field. 10%-20% of the research is high quality. In a 5-20 year range we'll discover exactly how enormous the true impact of commensal microbes on human biology is, even if the path is not a straight one and every publication is not correct. So science goes.
Student of another professor who was told to repeat his PhD experiment until the p-value was "acceptable".
Being published means very little in medicine.
Also, constant hype on HN for gut bacteria with credulous comments (or advertising, as another poster pointed out).