[R-sig-eco] Variance partitioning and sample size

[Melanie Croft-White]

Hi

I would like to answer the question "how much of the variance that we see
in species composition can be explained by the environment". We have 9
wetland sites that were sampled twice (in two separate years). Of the 9
samples 5 are in an impacted area and 4 are control.
After reviewing the literature I decided to use PCNM to partition the
variance but I don't know if we have enough sites to do this. We would have
18 sites if we used them as site-years. But then that incorporates temporal
variation into the mix of spatial and environmental variation. We would
also like to look at the difference in variation between the impacted and
control sites.
I haven't done PCNM before so I don't know if the same restrictions apply
as with other multivariate methods. From my experience I know that fewer
than 30 sites is kind of pushing it if you are doing a CCA. Is the same
true for PCNM? The papers that I have read that have used PCNM have sample
sizes ranging from 50-5000 (Dumbrell et al 2009; Jones et al 2008; Di Bie
et al 2012; Murphy et al 2015; Thompson et al 2012).
If anyone has any insights on sampling size I would greatly appreciate
their guidance and/or advice.

Thanks in advance,
Melanie

	[[alternative HTML version deleted]]