Hello,
I've used mantel tests a fair bit for standard genetic/geographic/landscape
analyses. I realize it involves permutations of rows and columns in the
matrix to obtain correlation coefficients.
However, for a certain question, I am interested in populations that are
nearest neighbours to each other, and identifying factors that affect
genetic distance between them. So because of autocorrelation problems, I
can't use standard regression or correlation, because some populations
appear several times depending upon their spatial arrangement. In essence I
am cherry picking a typical matrix, but only for populations that are
closest.
Is there code or a test that can handle this problem? It seems very similar
to a mantel, but only for a subset of possible pairs, not all possible
pairwise comparisons as a mantel would handle. What about a bootstrapping
algorithm than handles populations as the sample unit instead of individual
data points?
I bet mixed models could handle this problem somehow, but I'd like to stick
to the theme of a mantel test.
The data look like this
pop1 pop2 genetic_distance(dependent) (independent1)
popa popb data data
popb popd data data
popc pope data data
popf popg data data
popg poph data data
So for example, popb appears twice, as does popg. Clearly each row is not
independent.
Thanks for any insight
Rob.
--
View this message in context:
http://r.789695.n4.nabble.com/Mantel-like-analysis-tp3564296p3564296.html
Sent from the R help mailing list archive at Nabble.com.
