=== Training data set ===


=== Rating system ===

  * 1: Great cluster, tight and well isolated. The sort you would use as an example of what you'd like everything to look like.
  * 2: Very good cluster, a few points of contamination and/or touching edges
  * 3: Solid cluster, some contamination and slightly overlapping edges, but no major defects.
  * 4: Borderline cluster with some issues, e.g. diffuse, may be 2 cells, hard to separate in a principled way from neighbors.
  * 5: Cluster with defects. This is not included by default and will only be accessible when overridden (e.g. for decoding analysis)

By default ratings of 1-4 will be analyzed, and as an extra check only 1-3 can be done to help determine robustness of results to clustering errors.

Note: these are *not* relative ratings, i.e. if all your tetrodes are messy and there aren't any genuine "1s" then you don't give the best cells a "1" rating just because they are the best of a bad lot!