esantorella/hdfe

hdfe's Issues

Performance of Groupby

I found your blog post (http://esantorella.com/2016/06/16/groupby/) and was initially intrigued about being able to increase performance of user defined functions in groupby. But looking more closely I think you're just comparing a lazy function to a strict one. I don't see any difference in speed in terms of applying the functions to groups.

using the example from your blog, it think this shows most clearly what i mean:

In [30]: %timeit Groupby(df["first category"])
5.79 ms ± 103 µs per loop (mean ± std. dev. of 7 runs, 100 loops each)

In [31]: %timeit df.groupby("first category")
55.8 µs ± 3.99 µs per loop (mean ± std. dev. of 7 runs, 10000 loops each)

pandas groupby is lazy and does essentially nothing until you apply some function. your groupby does work in init to setup the groups.

Does that seem correct or am I missing how to speed up applying functions to groups?

This function always gives a syntax error.

Hi
I am trying to use this library but i am always getting a syntax error.

result[self.keys_as_int[k]] = function(vector[idx])
TypeError: only integer scalar arrays can be converted to a scalar index

Documentation

I read the article and code of your Groupby and really liked it. I am a bit lost with the other functions, but if Groupby is any indication, it would be worth learning about them. Could you write a documentation for them, similar to Groupby? At least what they are about and what the basic principles behind them are?

numpy function parameters

Hi
I want to use np.std over groupby and want to change the default value of ddof used in np.std.
Is this possible.

Thank you

Will groupby work if i have a list of columns to group?

This seems not to work?