Recommended setup involving Scitools, NumPy, and SciPy
Scipy/Numpy is the defacto standard for scientific/numerical computing with python. The vast majority of packages are built on top of them (including Scitools). In many ways it looks like Scitools is just a connivence wrapper around Numpy/Scipy/Matplotlib.
As far as Python 3 support, Numpy and Scipy are there, but Matplotlib is not yet (although I think it may be supported in the development repo on github):
http://onpython3yet.com/packages/requirements?r=numpy%0D%0Amatplotlib%0D%0Ascipy
Personally I've never used Scitools, since I prefer the lower level control of using the other libraries directly. I also stick with Python 2.7 since the complete scientific programming stack hasn't finished its migration to Python 3 yet.