[rprospero]

Numpy/Scipy are used for serious applications (I use them myself), but I think that the domains are different enough that the comparison is slightly unfair. Comparing Numpy to R for statistical analysis is like comparing R to Rails for web development.

Numpy isn't a statistics package. It's a linear algebra package. It's less of a poor-man's R and more of an easy-to-read APL.

As for the complexity and documentation, I honestly don't know where he's coming from in that regard. I've taught numpy to a variety of scientists in the department and they've always found it quite simple. The documentation is also thorough enough that these same people can use it to answer their own questions.