I was wondering if it is appropriate to use Bray-Curtis for dissimilarity calculations when you have an OTU table of rank abundance, i.e., OTUs for each sampling site are ranked from 1- OTUtotal in terms of abundance? (Raw sequence counts are not meaningful in my case.) Or would something like Gower distance be better? I would prefer to use Bray-Curtis because it excludes mutual zeros. When I use modified Gower distance (which also excludes mutual zeros) I get strange results (i.e., much higher stress in NMDS with a large NMDS axis 1 range). My worry with Bray-Curtis is that rank abundance is on an ordinal/interval scale instead of a continuous scale. Thank you!
- Alyse
I can suggest Gower’s dissimilarity coefficient, resulting in a species/species dissimilarity matrix. Gower’s metric is the most suitable for matrices
composed of continuous, ordinal and categorical variables.
While in small part the answer is that it depends on what you are trying to do, generally Bray-Curtis should be fine, either on the ranked abundances or on presence/absence (which becomes Sorenson/Dice). I wouldn't use Gower for a number of reasons, it's a distance so treat zeros as real, and it attempts to down-weight variables by their range, which in this case would be their maximum, and ends up standardising them (see Clarke KR, Somerfield PJ, Chapman MG (2006) On resemblance measures for ecological studies, including taxonomic dissimilarities and a zero-adjusted Bray-Curtis measure for denuded assemblages. J. Exp. Mar. Biol. Ecol. 330:55-80 for full expose on resemblances, available on my page).
- Badges
- Science topic
- Similar topics
- Medicine
- Physiology
- Biosignals
- Biomedical Signal Processing
- Statistical Analysis