I want to create a new distance matrix from an array of hundreds of distance matrices. This new matrix will be the average distances given a set of pairwise distances. The values to be averaged is based matching name prefixes. Easier to illustrate as my language might be incorrect: Here’s one sample distance matrix:

                    SpeciesA_1  SpeciesA_2   SpeciesC_1.  SpeciesC_2
SpeciesA_1      0       0.2         0.3     0.4

SpeciesA_2      0.2     0           0.4     0.5

SpeciesC_1      0.3     0.4         0       0.1

SpeciesC_2      0.4     0.5         0.1     0

I want the mean of all possible pairwise combinations given two prefix. For example:

SpeciesA_1 & SpeciesC_1 (0.3)
SpeciesA_1 & SpeciesC_2 (0.4)
SpeciesA_2 & SpeciesC_1 (0.4)
SpeciesA_2 & SpeciesC_2 (0.5)

Therefore, my resulting new distance matrix in this example should be:

           SpeciesA     SpeciesC
SpeciesA    0          0.4

SpeciesC    0.4         0

My actual matrices have about 140 columns and rows. I have no idea where to even get started with this. These dists are an array:

dists[,,i]

I can’t use a loop (which I would normally start with by selecting the first name) because it is an array which I never work with.

dim(dists)
[1] 135 135 138