Hello, I’m reopening this topic because I have a question related to the PyMC implementation of Stats Rethinking here: https://github.com/pymc-devs/pymc-resources/blob/main/Rethinking_2/Chp_14.ipynb.
If I understand the API correctly, PyMC’s implementation of the GP model should be calculating the “distances” automatically based on X. Correct? If yes, I’m not sure why the distance matrix is being passed into gp.prior:
gp.prior("k", X=Dmat)
Moreover, since the island data is using a regular coordinate system, wouldn’t we want to set input_dim equal to 2? Simultaneously, we’d want to replace the distance matrix with a matrix with shape (N, 2) where each row represents the location of the island on the x,y-coordinate system.
The way it’s currently implemented, aren’t we generating the prior based on distances between distances, that is, the distances between the distances between the first island and all the other islands?
What am I missing here?