0

I have species relative abundance data (as percentages) and several environmental parameters- and I have done normality tests on my data and it all seems to be normally distributed, but do I need to log transform the data anyway? I saw an online tutorial for CCA and it said to, but I would like to be sure.

kjetil b halvorsen
  • 63,378
  • 26
  • 142
  • 467
Jessica
  • 1
  • 1
  • 4
    Abundance as percent is at best roughly normal, if only because of the bounded scale. Environmental variables [in statistical terminology **not** parameters] are more commonly right-skewed than normal. But we can't be sure and you can't be sure whether transforming some or even all of your data is needed. You are likely to be asked, at a minimum, whether you considered transformation so the simplest advice is to consider analyses based on untransformed and transformed data and see which helps more. In any case, why presume that each variable must be transformed in the same way? – Nick Cox Oct 08 '14 at 10:59
  • Hi Nick, truth be told, I don't know much bout stats. I only thought I had to transform the data because a colleague showed me how to do CCA and transformed her species abundance data (but not the env variables). Plus some online tutorials say to transform it, but I wasn't sure. So do you mean I should run the CCA both ways and see which one looks best? – Jessica Oct 08 '14 at 11:09
  • 2
    We all need to do basic reading in texts and review papers if we want to understand something properly. You have summarized my advice accurately, except that it seems entirely possible to me that abundance data and environmental controls may need different treatment. – Nick Cox Oct 08 '14 at 11:27

1 Answers1

1

CCA is sensitive to outliers and assumes species response is a symmetrical unimodal function of position along environmental gradients. Hypothesis testing is based on randomization, so does not have distributional assumptions. But, CCA or not, transformations should be applied only if they improve data distribution (demonstrated using normality tests or PPCC fit).

katya
  • 2,084
  • 8
  • 11
  • 3
    "CCA or not, transformations should be applied only if they improve data distribution": that is too simplistic to escape comment. Many threads here show the use of transformations for linearising relationships, reducing inequality of scatter, removing or reducing trends, etc. In fact, it is arguable that transformation towards normality or Gaussianity is the least important kind of transformation, nor is formal testing particularly important or useful. See e.g. http://stats.stackexchange.com/questions/118214/r-test-normality-of-residuals-of-linear-model-which-residuals-to-use for some views – Nick Cox Oct 10 '14 at 21:53