One of my colleagues has sampled fish species from 5 different fresh water systems annually for a year. She also has physico-chemical measurement for each aquatic habitat for each month.
I think DCA would be an appropriate method to describe fish abundance among 5 habitats as well as each single habitat.
The propose here is of course description of fish species among five sites monthly at first. Another objective is testing of distribution pattern for each physico-chemical parameter.
She organized data in the form of month by species prior to analyze. Analyses were performed on vegan under R.
I shall be happy if I can get comment on analyses methods and methods that I suggested to use.
Thank you.
Chitra
There are two ways I think you can go here and they will somewhat depend on how well the data conform to assumptions of different techniques. Correspondence analysis can be effective if species tend to have a unimodal distribution across sites and all the usual assumptions are met. However, many times CA results in an "arch effect" that requires the use of DCA. The problem I have with DCA is that the detrending algorithm makes it difficult to know what the resulting bi-plot really means. For a an indirect gradient analysis I would go with NMDS. This technique is based on distances between samples in multivariate space and does not require the data to meet the same assumptions of CA or DCA. You can then either indirectly interpret the effect of environmental variables on the ordination or use sample scores on the axes of interest as a response variable in regression models to test for the effect of environmental parameters. If your data do meet the assumptions of CA/DCA, you can use canonical correspondence analysis (CCA). This is a direct gradient method that will ordinate species and sample scores along environmental gradients preventing the need for the two step process described above. All of these are available in "vegan". Good luck!
- Badges
- Science method
- Similar topics
- Biological Science
- Ecology