我一直在玩月末的不同定义，包括实际上一个日历日之前和之后的天数。熊猫有一些很好的工具，允许选择月末和使用shift（）方法我能够在该月的实际最后一天之前和之后的几天移动日期。
我苦苦挣扎的是选择每月间隔排除那些日子在月末的日子，并最终写出dict理解的组合并调用pd.Series。我确信有更好的方法来做我做的事情，我很乐意找到它。以下是供参考的代码：
%matplotlib inline
import seaborn as sns
import pandas as pd
from datetime import datetime
from pandas.io.data import DataReader
start_date = '2012-09-09'
end_date = datetime.today()
SPX = DataReader('^gspc', 'yahoo', start_date, end_date)
#Select the month end straddle interval
back, fwd = -3, 4
me_range = abs(back) + fwd
#Calculate dates corresponding to the month end interval
dt = [pd.date_range(start_date, end_date, freq='BM').shift(i, freq='B')
for i in range(back,fwd)]
#Filter time series for rest of the month values
rest_of_month_data = SPX.Close[SPX.index-dt[0].append([dt[i] for i in
range(1,len(dt))])]
#Extract values for the last day of the month before the beginning of month end range
last_day = pd.Series({rest_of_month_data.index[i-1] : rest_of_month_data[i-1]
for i in range(1,len(rest_of_month_data))
if rest_of_month_data.index.month[i] !=
rest_of_month_data.index.month[i-1]})
last_day = last_day.append(pd.Series({rest_of_month_data.index[-1] :
rest_of_month_data[-1]}))
#Extract values for the beginning of the month after month end range
first_day = pd.Series({rest_of_month_data.index[j] : rest_of_month_data[j]
for j in range(len(rest_of_month_data)-1)
if rest_of_month_data.index.month[j] !=
rest_of_month_data.index.month[j-1]})
#Calculate returns
rest_of_month_rets = pd.Series({last_day.index[i]: 1 + (last_day.values[i] -
first_day.values[i]) /
first_day.values[i]
for i in range(len(last_day))}).cumprod()
#Plot results
(rest_of_month_rets - 1).plot(title = 'Rest of Month Strategy Returns');
我试图通过不属于dt的值（我分配给月末范围的变量）来索引我的系列，但由于我保留了AttributeError，我无法出于某种原因：'NotImplementedType'当我试图运行时，对象没有属性'dtype'：
ES.Open[ES.index!=dt[0].append([dt[i] for i in range(1,len(dt))])]
选择所需的日期。这就是我使用之前编写的函数的原因。为完整起见，这是完整的错误消息。
---------------------------------------------------------------------------
AttributeError Traceback (most recent call last)
<ipython-input-38-aa2d213a3d00> in <module>()
----> 1 ES.Open[ES.index!=dt[0].append([dt[i] for i in range(1,len(dt))])]
/usr/local/lib/python3.4/dist-packages/pandas-0.15.1_91_g0a2ea0a-py3.4-
linux-x86_64.egg /pandas/tseries/index.py in wrapper(self, other)
79 elif not isinstance(other, (np.ndarray, Index, ABCSeries)):
80 other = _ensure_datetime64(other)
---> 81 result = func(np.asarray(other))
82 result = _values_from_object(result)
83
/usr/local/lib/python3.4/dist-packages/pandas-0.15.1_91_g0a2ea0a-py3.4-linux-
x86_64.egg/pandas/core/index.py in wrapper(self, other)
55 # technically we could support bool dtyped Index
56 # for now just return the indexing array directly
---> 57 if com.is_bool_dtype(result):
58 return result
59 try:
/usr/local/lib/python3.4/dist-packages/pandas-0.15.1_91_g0a2ea0a-py3.4-linux-x86_64.egg
/pandas/core/common.py in is_bool_dtype(arr_or_dtype)
2459
2460 def is_bool_dtype(arr_or_dtype):
-> 2461 tipo = _get_dtype_type(arr_or_dtype)
2462 return issubclass(tipo, np.bool_)
2463
/usr/local/lib/python3.4/dist-packages/pandas-0.15.1_91_g0a2ea0a-py3.4-linux-x86_64.egg
/pandas/core/common.py in _get_dtype_type(arr_or_dtype)
2388 elif isinstance(arr_or_dtype, CategoricalDtype):
CategoricalDtypeType
-> 2390 return arr_or_dtype.dtype.type
2391
2392
AttributeError: 'NotImplementedType' object has no attribute 'dtype'