On Tue, 12 Jan 2021 09:43:35 +0000
Constanze Keye <constanze_keye at hotmail.com> wrote:
Dear list members, I struggle over a few things whilst doing a RDA for the last couple of days. Could please somebody enlighten me? 1. Do I have to standardise my explanatory data before doing a RDA (function rda, vegan)? 2. I know that_s a normal procedure before RDA, but I found this post in the r-sig forum: https://stat.ethz.ch/pipermail/r-sig-ecology/2010-April/001196.html Now, I'm not sure if I can/should skip standardisation before RDA using the vegan package, or if I did understand something wrong here? 2. Why are all my species always clustered in the very middle of my RDA plot ? If I use function 'triplot.rda' from https://raw.githubusercontent.com/zdealveindy/anadat-r/master/scripts/NumEcolR2/triplot.rda.R , they are much more spaced around the plot and some of them seem to correlate well with my explanatory variables... Sorry, if these questions are a bit basic, but I_m still trying to get my head around the whole subject of RDA. Cheers, Conny [[alternative HTML version deleted]]
Dear Conny, regarding to the second point, I think it is crucial to "zoom in" the cluttered part - if the pattern is the same (or just flipped), then everything is OK, you just want to rescale your explanatory variables' scores to get them into the plot. Various SW packages use various scaling, choices of the vegan authors are described here: https://cran.r-project.org/web/packages/vegan/vignettes/decision-vegan.pdf But - and I hope J.O. or someone else will tell me if I am wrong - if you use those explanatory variables as "directions" (RDA belongs to "linear world", so there are no peaks of explanatory variables, just directions in which they increase), you can scale their scores with any reasonable constant (*constant*, so keeping the ratios), thus effectively zooming the plot to contain both the site "points" and explanatory variables as "arrows". HTH, Martin Weiser