- Datasets
- A dataset is the samples one study published. We analyse them ourselves, so a count of datasets isn’t a count of what those studies concluded. Only datasets that had people with the condition and healthy people, 10 or more of each, are used. Numbers come from those datasets, never from pooling.
- How common
- The share of people whose sample has any of it, in the people with the condition and in the healthy people of those same datasets. Tested within datasets (Cochran–Mantel–Haenszel, corrected for testing many microbes); a difference counts when it is clear and meaningful (Cohen’s h of 0.2 or more).
- Strength word
- Counted by code from how many datasets point the same way: Seen again and again (3 or more, none clearly the other way), Seen more than once (2, the majority), Seen once (1, listed last). Where datasets disagree, a row isn’t listed.
- Each dataset
- Tap a row for every dataset’s own numbers, and for how much carriers have.
