PathAverage interspecimen mean

Top image shows raw data. left is after quantising. Right is using gaussian smoothing with 5 elements.

Don't think it makes sense to add smoothing when we average.

In orange we have interspecimen average that keeps NaN => fillmissing, and in purple we have fillmissing => mean.
The features are the same, but average => fillmissing is smoother.

Start by locating the MATLAB implementation of PathAverage interspecimen averaging and inspect how fillmissing, averaging, and any Gaussian smoothing are ordered. Compare that behavior with the two cases described in the issue; done should mean the intended NaN handling and smoothing behavior are explicitly decided and verified against the reported plots.

Written by the indexing model from the issue text.