[R-sig-eco] adonis with continuous variables

[Jari Oksanen]

Hello Alfredo,
On 10/03/11 01:11 AM, "alfredo Tello" <alfredotello at gmail.com> wrote:
 
> The adonis documentation in vegan says that adonis accepts categorical
> factors and/or continuous variables. I was aware (have seen it posted
> somewhere else) that adonis could be run using a continuous variable as a
> co-variate, but not that it could (or was initially intended to) be used
> only with continuous variables.
>
Indeed, adonis() can use only continuous variables, it is documented to do
so, and this was initially intended.
 
> Is it appropriate to use it with a single or several continuous explanatory
> variables in a regression type approach? I have noticed that I get very
> similar results with db-RDA (capscale) when using both functions with
> continuous variables.
> 
Yes, this is adequate. The results are expected to be very similar to dbRDA,
and with Euclidean distance they should be identical. Imaginary dimensions
(negative eigenvalues) are handled differently in adonis() and in capscale()
and this causes some differences, depending on the magnitude of the negative
eigenvalues.

Cheers, Jari Oksanen