Change default column names in read_* functions when header=False

[wesm]

In retrospect I think using R's convention of X.1, ..., X.N was a bad move. Maybe range(N) would be better? Will cause API breakage but not the end of the world

[wesm]

@changhiskhan do you have an opinion?

[jseabold]

FWIW, I don't like the dots because you lose attribute access, but I do like the X. I believe we grab and use the column names in statsmodels.

[changhiskhan]

I think the API breakage will be worth it in exchange for attribute access.
Maybe just X1 .... XN or X_1 ... X_N?

[wesm]

Or start at 0?

[jseabold]

Maybe more Pythonic to start at 0, but I don't think people really think of DataFrame columns positionally. No strong opinion either way.

[changhiskhan]

I vote for starting at 0 since there's no compelling reason to break consistency here.
Otherwise it's a little weird that DataFrame(data) starts numbering the columns at 0 but pd.read_csv('frame.csv') gives columns that start at 1.

[wesm]

OK. I'll bite for 'X0', ... , 'XN'. Hopefully will not cause too much damage