Actually, I think there are a few other bugs I didn't address here:

1. I believe new_distance_gpa is defined incorrectly... this should be a measure of the distance between the individual shapes and the mean, but currently measures the change in the mean from one iteration to the next. I think the correct version is something like the following, assuming what we want is the mean distance between individual shapes:

   new_distance_gpa = np.mean([np.square(arr - new_ref).sum() for arr in array_list], axis=0)

2. I think a better convergence criterion to use is something like np.linalg.norm(ref - new_ref) < tol... i.e., we want to check if the reference shape is stable, rather than if the distance is stable.

Codecov Report

All modified and coverable lines are covered by tests ✅

Please upload report for BASE (main@d4f2ab6). Learn more about missing BASE report.

@@           Coverage Diff           @@
##             main     #217   +/-   ##
=======================================
  Coverage        ?   94.13%           
=======================================
  Files           ?       11           
  Lines           ?      802           
  Branches        ?        0           
=======================================
  Hits            ?      755           
  Misses          ?       47           
  Partials        ?        0           

Hi @msegado . Thank you for noticing the issue. I will take a closer look at the PR during the weekend.

Thanks for noticing the issue in the codes. I am going to merge this PR and add tests later. As for two questions you mentioned, I will open issues and address them. @msegado