I am currently running a sPCA on GBS data from a number of plant populations following the tutorial provided by Thibaut Jombart.

I understand one important decision to make during the analysis is the choice of a connection network. However, I have been unable to find literature on what might make a certain type of network suitable for some type of organism or not (I am working on a herbaceous perennial). Is someone aware of some publication discussing this or has some experience with the method? Any insight is greatly appreciated.

What I can say for my data set so far: The general geographic pattern is always recovered, unless I chose method=5 (neighborhood by distance) which yields a rather different structure (choosing fairly random min and max distances).