I am currently looking at whether populations are more similar in their sensitivities to climate if they are geographically closer to one another. I would expect that populations that are further apart might have fairly different sensitivities compared to those that are closer together.
But how to test this?
The problem is that the data are not independent. This is because the difference in sensitivity between population 1 and 2 uses some of the same information as the difference between populations 1 and 3. This means that if you were to carry out a normal regression on the data the standard errors would be much smaller than they actually are, so you are more likely to get a “significant” result where you might not actually have one.
Mantel’s test
This analysis is used for testing null hypotheses about correlations between matrices. In order to get around this issue of non-independence, the data are randomised and then the real data is compared to it. This tells you whether the relationship between two matrices (in my case difference in sensitivities and distance) is more different than we would expect by chance.
The main problem with the Mantel test (at least for me) is that it is only based on linear correlations, so nonlinear relationships can’t be investigated.
I will have to keep you posted on the answer!
The main problem with the Mantel test (at least for me) is that it is only based on linear correlations, so nonlinear relationships can’t be investigated.
I will have to keep you posted on the answer!