Hello there!
Im working with a community matrix (abundance data), and i would like to check if there is a taxa indicative of different trophic state of lakes. I would to that with Indval command in R.
The problem is that my data has zeros and some taxa abundance goes up to 10000 individuals per sample.
In this scenario, would it be advisable to transform my data prior to the analisys? Should i use a Hellinger transformation?
I ask about a hellinger transformation because im thinking of doing after a CCA in this same community matrix, with some environmental data.
Any recommendation?
Thank you for your time!
You should definitely transform your abundace data. Also consider the Winsconsin double transformation and a non-metric multidimensional scaling (function meta_MDS in vegan) instead of a cca.
Hope this helps
Thank you Facundo! Ill check these methods that you recommended!
Hi, according to my experience, it is reasonable to use Hellinger transformation.
Lucas, Hellinger transformation is meant to solve the issue of double absences when computing Euclidean-based ordination techniques (PCA, RDA). Thus, it is not appropriate for CCA. Now, in the context of indicator species, you have to ask yourself if transforming your data (Hellinger or otherwise) is meaningful, especially if same sampling effort is put into collecting species from each of the categories you want to find indicators for. Remember that the indicator value doesn't depend only on the relative abundance of the species among categories but also its relative frequency, and the fact that a species has much higher abundances than others in the system must have some ecological implications. Now, the indicator value of one species is independent from the value of the other species in the assemblage, thus transforming for the purpose of accounting for large differences in abundance has little use. If you apply Hellinger (the square root of the relative abundance among species within a sample) and then compute the indicator value (highest value among sites of the product of the relative abundance and relative frequency of a species within a site) you are no longer values independent among species. Hope this helps.
Hi Lukas,
the two things of IndVal and CCA are unrelated. Regarding IndVal, I would suggest you NOT to transform the data, simply because the mean abundances across the groups would not be the real one. Moreover, as mentioned by Jaime Pinzon before, if you have striking abundance of a particular species/taxon, it will not be a detrimental issue for the other species (when you look for indicator ones).
- Badges
- Science topic
- Similar topics
- Mathematics
- Algebra
- Matrix