Mantel between similarity and dissimilarity distance matrices

[fried1]

Hello,

We are trying to decide whether or not a result that we're finding to be significant is truly so. We have two matrices that we're testing, one is derived from Bray-curtis similarity, one is derived from Gower dissimilarity. Is is appropriate to interpret a significant p-value in our case, given that we're not testing two matrices of the same "type"? Would it be more appropriate to convert our B-C matrix to a dissimiarlity matrix by subtracting all of the values from 1?

Thanks.

[ohammer]

You have two options:

1. Keep the matrices as they are, but remember that the correlation (R) value will now be negative for a positive correlation of similarities, so make sure that the p value is computed using the negative tail (or use the two-tailed significance).

Or 2: As you say, just convert e.g. the BC values to 1-BC.

I would prefer the latter - seems more logical.

[bugman]

Or, if you are using PRIMER, you can do this using the RELATE routine.

[ohammer]

Or, if you are using Past, if you mix similarity and distance measures then the Mantel test module will automatically adjust the test accordingly.