0

I have a dataset with presence-absence species data measured at several different sites. The data was measured over the span of 10 years. On many sites, measurements were taken several times within a year. The frequency of measurements is not constant nor were all sites measured several times, some were even only measured once.

I know that a classical Detrended Correspondance Analysis is not helpful here, since it does not consider the cofactor time. Is there any way to include all sampling points or any other correspondance analysis method that is useful here?

Thanks a lot for any help!

gamel
  • 29
  • 6

1 Answers1

1

If you want to estimate the time effect or partial it out, yes, but not in vegan. Canoco has detrended canonical correspondence analysis (DCCA), the constrained form of DCA but vegan doesn't and is unlikely to ever have it.

There's nothing stopping you throwing all samples into a DCA you just can't remove the temporal effects.

Alternatively, choose a suitable dissimilarity coefficient and use NMDS via vegan's wrapper metaMDS(). This will give you a DCA-like analysis. If you want to account for the temporal effects, then using the same dissimilarity look at dbrda() as one option.

Gavin Simpson
  • 170,508
  • 25
  • 396
  • 453
  • **vegan** will never have DCCA. However, DCA is just an exploratory technique, and nothing stops you from throwing any non-negative data into DCA. – Jari Oksanen Jan 11 '18 at 08:27