Message-ID: <CAPQ=hV+FrMYXJuPAbwok1fV8xcB+9DayVwSf4mfZ-oh2105QbA@mail.gmail.com>
Date: 2012-02-01T17:48:06Z
From: Oliver Soong
Subject: vegan::adonis, SS, order of terms
In-Reply-To: <C8D016E3-41CB-4304-B4DC-7EE496D3BF2B@oulu.fi>
Thanks for reminding me about the behavior of stats::anova under
unbalanced designs. I had forgotten that equations 5 and 6 in
Anderson 2001 are only for balanced designs.
Oliver
On Wed, Feb 1, 2012 at 2:40 AM, Jari Oksanen <jari.oksanen at oulu.fi> wrote:
>
> On 01/02/2012, at 06:52 AM, Oliver Soong wrote:
>
>> Are the vegan::adonis sums of squares supposed to depend on the order of the
>> terms?
>
> Yes, the terms are added sequentially, just like in ordinary R anova, and they are order-dependent. The next release version of vegan will show this information (I just committed the change to R-Forge).
>
>> I'm trying to understand permutational MANOVA, and this
>> disagrees with my reading of Anderson 2001.
>>
>> library(vegan)
>> data(dune)
>> data(dune.env)
>> adonis(dune ~ A1 + Management, dune.env)
>> adonis(dune ~ Management + A1, dune.env)
>>
> So *how* does this disagree with your reading?
>
> Cheers, Jari Oksanen
> --
> Jari Oksanen, Dept Biology, Univ Oulu, 90014 Finland
>
>
>
>