I have recently changed the display below to show the percentage of matched instead of just the number of bacteria matches (the number will appear if you hover over the link as a tool tip). The numbers may be prone to misinterpretation, hence this technical page.
The candidate bacteria comes from special studies — it is important to note that often these bacteria are rarely seen, so having a 100% match is effectively impossible. We also have the dilemma of a single sample versus a collection of samples.
The rule that I am using is simple, a match must:
- Have the bacteria (if it is missing, it is not deemed a match)
- The bacteria count must be either:
- below the study mean – 3 standard deviations of the mean if the study found it to be a low mean value against the reference population
- above the study mean + 3 standard deviations of the mean if the study found it to be a high mean value against the reference population
Naively, assuming a normal distribution, the odds of a single match is around 1%, so with 200 items to check, we would expect 1% for a random person.
You should NOT view these as predictive, for example both ME/CFS with IBS and ME/CFS without IBS are on the list with the same value!!! Instead, your existing condition(s) should be used to select only the ones that apply to you. You could arbitrarily do all of the high ones — I do have a concern about that approach, you are creating noise that may make suggestions less effective.
One last item is the quality of the read (i.e. how many bacteria was actually detected in the sample). Since we are dealing with rare bacteria, bacteria (that are actually there) may not be detected and thus you have a lower percentage match. So do not view the percentage as absolute. but relative to others in the sample.