Accommodate DataFrames with numeric column names

As demonstrated in the example from [this stack overflow response](http://stackoverflow.com/a/42164577/3585557), when converting a NumPy array into a Dataframe, the column labels are initially numeric. When trying to index by these column labels, I realized the backend DataFrame indexing is using the attribute method, `df.column_label`, and not the `df[column_label]` method. The current method supports less variation in column label in that it only supports valid python identifiers, e.g., column labels cannot be purely numeric. [This Pandas' doc page](http://pandas.pydata.org/pandas-docs/stable/indexing.html) has more details. If possible, changing to the `[ ]` method (or `__getitem__ method`) would allow for a wider range of inputs, thereby simplifying the user experience.