Probability is an estimate whether something may help, not how much it can help. The relative help between two items is rarely found in any study.
Probability is based on the number of studies reporting that something shifts a bacteria in the desired way. A single report that Blue Cheese reduces Xeonella may get a value of 0.1 while 10 reports that barley reduces Boozella would get a value of 1.0 (if 10 reports were the maximum number of reports reported).
Studies often contradict each other – typically caused by a confounder that the study ignored.
https://en.wikipedia.org/wiki/Confounding. To address this, we aggregate the number of reports with scaling. For example:
- 6 reports showing desired changes
- 2 reports showing undesired changes
Could be computed as 6-2=4. Due to this increase uncertainty, we do other methods for example:
- exp (log(6/8) + log(2/8))
Once we compute all of these numbers, we then scale them so the maximum value is +1.