Pandas: Feature request: add median, mode & number of unique entries to pandas.DataFrame.describe()

[tantrev]

Would come in handy for a number of a different applications from basic statistics, to understanding one's data, to estimating machine learning algorithmic load.

dupe of #2749

[jreback]

FYI, you can easily make your own and patch it in. Just put in your startup code / application.

def describe(self):
    """ describe  of a series """
    l = [ ('nobs'  , len(self.index)),
          ('valid' , self.count()   ),
          ('mean'  , self.mean()    ),
          ('min'   , self.min()     ),
          ('max'   , self.max()     ),
          ('std'   , self.std()     ),
          ('10%'   , self.quantile(0.10)),
          ('25%'   , self.quantile(0.25)),
          ('50%'   , self.median()  ),
          ('75%'   , self.quantile(0.75)),
          ('90%'   , self.quantile(0.90)),
          ('skew'  , self.skew()    ),
          ('kurt'  , self.kurt()    ) ]
    s = Series(dict(l), index = [ k for k, v in l ])
    s[s.abs()<0.000001] = 0.0
    return s
Series.describe = describe

[soupault]

median out-of-the-box would really be a great thing to have!
Not just for Series but also for GroupBy's.

[jorisvandenbossche]

Note that median is already in describe, as in 50%

[soupault]

@jorisvandenbossche indeed! Sorry, my bad. :baby:

[TomAugspurger]

Hijacking this issue to propose a variant (and steal more from dplyr)

Why not have a DataFrame.agg that is identical to DataFrame.groupby.agg, but with a "single group"? This also goes along with the new .resample.agg, yay for synergy. Basically it takes a function/str or list of functions/strs or a dict of column names to functions/str and aggregates accordingly.

[jorisvandenbossche]

Related to https://github.com/pydata/pandas/issues/1623