Python dataframe - 每年有条件的平均

Question

I have the following code:

import csv
import pandas as pd
import numpy as np


stocks_dataframe = pd.read_csv('^GSPC.csv', delimiter = ',')        


stocks_dataframe['Percent_change'] = stocks_dataframe['Close'].pct_change()
stocks_dataframe['positive_return_day'] = np.where(stocks_dataframe['Percent_change']>=0, 1, 0)
stocks_dataframe['negative_return_day'] = np.where(stocks_dataframe['Percent_change']<0, 1, 0)
stocks_dataframe['positive_return_day'].value_counts()
stocks_dataframe['date'] = pd.to_datetime(stocks_dataframe['Date'])
stocks_dataframe['year'], stocks_dataframe['month'] = stocks_dataframe['date'].dt.year, stocks_dataframe['date'].dt.month
yearly_data = pd.DataFrame()
yearly_data['positive_return_day'] = stocks_dataframe['positive_return_day'].groupby([stocks_dataframe.year]).agg('sum')
yearly_data['negative_return_day'] = stocks_dataframe['negative_return_day'].groupby([stocks_dataframe.year]).agg('sum')


stocks_dataframe.groupby(stocks_dataframe.year)['Percent_change'].transform('mean')

How can I calculate the average return separately for positive return days and negative return days? I would like to get these values per year and store them in the yearly_data dataframe.

Here is the head of the stocks dataframe:

stocks_dataframe.head()
Out[35]: 
         Date         Open         High  ...  year  month  negative_return_day
0  1999-12-31  1464.469971  1472.420044  ...  1999     12                    0
1  2000-01-03  1469.250000  1478.000000  ...  2000      1                    1
2  2000-01-04  1455.219971  1455.219971  ...  2000      1                    1
3  2000-01-05  1399.420044  1413.270020  ...  2000      1                    0
4  2000-01-06  1402.109985  1411.900024  ...  2000      1                    0

[5 rows x 13 columns]

Answer 1

can't you just groupby again?

for year, df in stocks_dataframe.groupby(stocks_dataframe.year):
     print(year)
     print(df.groupby(df.negative_return_day).Percent_change.mean())

edit: now you can get the year too