#this polyfit script is modeled after polyfit-new.py, it takes the new csv file with 30point data vector, and tone attribute for Xu's data #Here we adopt a simpler design, not to use column wise data ...

np.ma.polyfit fails if there is a whole column of missing values. I think some broadcasting is incorrectly handled under the hood. See MWE for clearer description. Note that regular np.polyfit works ...