We have a fish juvenile abundance dataset with a lot of zeros (the catch was low), so wen trying to get a resemblance matrix, the Bray-Curtis similarity index does not run well (the resulting matrix has a lot of "undefined" values), but Euclidean distance works well. Can I use Euclidean instead of B-C? or what can we do to solve this problem? We are trying to run multivariate analyses. Thanks!
Hi Macarena!, I'm not specialist with Multivariate analysis, but what I have found in other ocasions sounds like this:
If you want to treat the data as presence/absence data, then Jaccard coefficient (also called the coefficient of community) and the slightly modified Sorensen coefficient (also called the Dice index) can be used (Legendre and Legendre (1998)). If you want to take into account the quantitative aspect of the data you can use the Similarity Ratio (SR) and the percentage similarity index (also called Bray-Curtis coefficient and the Czekanowski coefficient).
Then you should use appropriate analysis like principal co-ordinate analysis (PCoA), non-metric multidimensional scaling (NMDS), the Mantel test or ANOSIM. For ordination and clustering purposes, the Sorensen or Chord distance functions seem to perform well.
I hope this info can be useful for you!,
- Badges
- Science topic
- Similar topics
- Indexes