> There should be more overlap between NetworkX and NumPy, than with Pandas, esp among Spectral Graph Theory practicioners Sounds partly like sampling bias to

> There should be more overlap between NetworkX and NumPy, than with Pandas, esp among Spectral Graph Theory practicioners Sounds partly like sampling bias to me. I do not at all think it's safe to assume that GitHub users that star libraries actively is a representative random sample of all users of those libraries. But it also sounds like a misunderstanding of just how many people think of "Numpy + Pandas" together as forming the foundations of the data analysis and data processing toolkit in Python. Almost every project where I use the latter, I end up importing the former for some reason or other, even if I'm not doing anything heavily numerical or computational.