I have a bacterial OTU matrix from various sample points and times. So for each sample I have the sequence counts across the various OTUs that they were assigned to.

I also have a corresponding matrix of abiotic variables including temperature, pH and the concentrations of a range of nutrient compounds. This data was collected at the same time as the bacterial OTUs.

This is a tutorial written by the people behind the R package 'vegan', it offers three separate types of constrained ordination. I just don't understand what the differences are between these and what would suit my data-set best. There is not much advice out there at all for this sort of thing so any help would be appreciated.


1 Answer 1


You might want to check out Mike Palmer's webpage on ordination, specifically the constrained ordination section: http://ordination.okstate.edu/

The main difference between constrained ordination techniques is the distances that are preserved. Redundancy analysis (RDA) preserves Euclidean distances, so in order to use it, you need to transform your OTU table with an appropriate transformation (e.g., Hellinger), after which a Euclidean distance calculation is appropriate. It also works best identifying linear relationships between environmental variables and species responses. Canonical correspondence analysis (CCA) preserves the chi-squared distance, which tends to emphasize common taxa, and looks for unimodal responses to environmental gradients. Distance-based redundancy analysis (dbRDA) accepts any distance matrix, allowing you to use non-metric distances, and then looks for linear relationships with distances and environmental variables.

Before applying these metrics, you'll need to understand their advantages and limitations, and then you'll need to make a decision about which combination of constrained ordination techniques and distance metrics are best for answering your questions. This is something you'll need to justify yourself, based on your own question, because different approaches are better at emphasizing different aspects of your data. I strongly suggest you study a copy of Legendre and Legendre (2012) for more information about these methods.

  • $\begingroup$ There is also a good chapter about ordination in this book $\endgroup$
    – RHA
    Oct 12, 2017 at 17:53

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.