The Microbiome in Posttraumatic Stress Disorder and Trauma-Exposed Controls: An Exploratory Study

Paywalled at http://journals.lww.com/psychosomat...iome_in_Posttraumatic_Stress_Disorder.14.aspx

Article on paper - https://www.sciencedaily.com/releases/2017/10/171025103140.htm

 

random forest analysis? WTF? Is this some form of p-hacking or is there something more to it?

 

full text at scihub: https://sci-hub.cc/10.1097/PSY.0000000000000512

Let me share with you the author list for this study:

That lightbulb sure must have been tough to change, eh?

The primary result is that there were no significant differences between groups in microbiome diversity, contrary to their predictions.

The forest thing they did (Variable Selection using Random Forests), trawls through a large dataset to find variables that reliably distinguish your two groups. If you have a large enough set of variables, you are pretty much guaranteed to find something with this approach. So you can't really conclude anything from it. All you can use it for is to generate a hypothesis to test on your next set of participants. They also tried to do various confirmatory analysis of the variables after they had dug them out, but they've used correlation, which doesn't seem to me to be the right approach when you have two groups that are strictly delineated.

Still, at least they were honest about their predictions not being confirmed. They could have done what a lot of their colleagues done, and re-written the intro to make their predictions fit with the findings.