pandas rolling_min and rolling_max different behavior?

Question

I have a pandas data frame with three columns id,date,value.

Out[411]:
symbol  date_col    PX_HIGH
0   BF/B US Equity  2014-01-02  75.6800
1   DLTR US Equity  2014-01-02  56.5600
2   EMN US Equity   2014-01-02  80.5300
83620 rows × 3 columns

I would like to compute the rolling max and min on these values grouping by symbol. For rolling max

df.groupby('symbol')['PX_HIGH'].apply(lambda x: pd.rolling_max(x,window=20,min_periods=20)

And I get

Out[418]:
0    NaN
...
83605     54.5400
83606     56.0500
Length: 83620, dtype: float64

However, when I do the samething using rolling min

df.groupby('symbol')['PX_HIGH'].apply(lambda x: pd.rolling_min(x,window=20,min_periods=20)

I get

ValueError                                Traceback (most recent call last)
<ipython-input-419-030ef09b7d35> in <module>()
  1 
  ----> 2 update_df.groupby('symbol')['PX_HIGH'].apply(lambda x: pd.rolling_min(x,20,20))
  ...

  ValueError: min_periods (20) must be <= window (17)  

Any reason for why this might be? I'm using pandas 0.13.1

Answer 1

我不确定为什么pd.rolling_max()不能工作，但pd.rolling_min()不能工作，但我认为此问题已在Pandas pd.rolling_min()中修复-请参阅http://pandas.pydata.org/下的第一项pandas-docs / version / 0.15.0 / whatsnew.html＃whatsnew-0150-roll （ https://github.com/pydata/pandas/pull/7766 ）。