Dear Meredith,

It is absolutely OK to do constrained (aka direct) ordination with only one environmental variable. You can do it e.g. in vegan through RDA (using euclidean dissimilarity in species composition among samples - not recommended:), or through CCA (Chi-square dissimilarity) or dbrda (using any ecologically meaningful dissimilarity you want - this would be my choice).

Best regards,

martin

I agree with Martin, you could also display the dbRDA ordination describing the relationship between the environmental variable and the taxonomic dissimilarity between samples. Vector length corresponds to the relative size of the effect, whereas direction represents the correlations to the displayed axes. It will also show you the percentage the first axis captures of fitted and of total variation between samples.

Here is an example of how dbRDA is used in an article:

https://www.nature.com/articles/ncomms3457

Ordination with only one environmental variable is not very useful imo, except to show the size of the effect. So you also need to use regression analysis, but you might have to transform the % data :)

Hi Meredith,

Constrained ordination with only one explanatory variable is not the best choice. Yes, you can do it, and there is nothing wrong about it; however, as the main purpose of ordination in general is to reduce dimensionality, in this case you only get one ordination axis (despite the fact that you have several species) and thus you only can assess the gradient in one dimension. The alternative (and not sure how great it is) to 'force' the analysis to by converting your one continuous variable into a categorical variable. Of course, you'd need to have a good and sound rationale to assign categories to it (e.g., natural breaks or something alike). Alternatively, you can double-check if your two other variables are correlated (i.e., colinear) with the one you are interested in, and if not, you can still use them in the analysis. Whether they are significant or not in the model is another story...