Strata vs. Condition in vegan CCA

Hello all,

I am looking at a microbiome dataset of ~150 individuals from 11
populations. I have ~15 population-level covariates, and I want to compare
these covariates with microbiome composition while controlling for
population. My initial thought was that I would need to aggregate the
microbiome data by population before performing constrained ordination, but
I believe this would result in a loss of statistical power, and that
instead I should retain individuals while somehow controlling for
population.

If I understand correctly, I can control for population structure in
vegan::cca using either Condition(Population) as part of the cca model, or
using strata = Population when I run anova.cca().

I have two questions related to this analysis:
1. Is it technically possible and statistically valid to include a
conditioning variable that is fully confounded with the ~15 other cca model
terms? e.g. if all my cca terms are measured at the population level, can I
control for population to avoid pseudoreplication? Or does that leave me
with effectively zero degrees of freedom?
2. What are the theoretical and computational differences between using a
Condition(Population) in the CCA model versus including strata = Population
in anova.cca?

Thanks in advance for your input.

Claire


Claire E. Couch (she/her/hers)
NSF Graduate Fellow
Department of Integrative Biology, Oregon State University
--
*Oregon State University in Corvallis, OR is located within the traditional
homelands of the Mary's River  or Ampinefu Band of Kalapuya.  Following the
Willamette Valley Treaty of 1855 (Kalapuya etc. Treaty), Kalapuya people
were forcibly removed to reservations in Western Oregon. Today, living
descendants of these people are a part of the Confederated Tribes of Grand
Ronde Community of Oregon (https://www.grandronde.org
<https://www.grandronde.org/>) and the Confederated Tribes of the Siletz
Indians (https://ctsi.nsn.us <https://ctsi.nsn.us/>).*