In [4]:
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
import statsmodels.api as sm
from scipy.stats import chi2, f
import warnings
warnings.filterwarnings("ignore")
In [6]:
stock_data1 = pd.read_csv('1EShowData_data_5stocks_daily_price_2001-2010.csv',
encoding='GB2312',
usecols=[0, 1, 5, 10])
stock_data1
Out[6]:
In [7]:
stock_data2 = pd.read_csv('1EShowData_data_5stocks_daily_price_2011-2015.csv',
encoding='GB2312',
usecols=[0, 1, 5, 10])
stock_data2
Out[7]:
In [8]:
stock_data3 = pd.read_csv('1EShowData_data_5stocks_daily_price_2016-2018.csv',
encoding='GB2312',
usecols=[0, 1, 5, 10])
stock_data3
Out[8]:
In [9]:
stock_data = stock_data1.append(stock_data2, ignore_index=True)
stock_data
Out[9]:
In [10]:
stock_data = stock_data.append(stock_data3, ignore_index=True)
stock_data
Out[10]:
In [11]:
stock_data.columns = ['code', 'date', 'close', 'rfreturn']
stock_data.dropna(inplace=True)
stock_data.sort_values(by=['code'], inplace=True)
stock_data
Out[11]:
In [12]:
stk_codes = np.unique(stock_data['code'].values)
stk_codes
Out[12]:
In [20]:
stock_data50 = stock_data[stock_data['code'] == stk_codes[0]]
stock_data50['date'] = pd.to_datetime(stock_data50['date'])
stock_data50.sort_values(by=['date'], inplace=True)
stock_data50['return'] = np.log(stock_data50['close']) - np.log(stock_data50['close'].shift(periods=1))
stock_data50.dropna(inplace=True)
ind = (stock_data50['return'] >= -0.1) & (stock_data50['return'] <= 0.1)
stock_data50 = stock_data50.loc[ind, :]
stock_data50
plt.plot(stock_data50['return'].values)
Out[20]:
In [21]:
stock_data51 = stock_data[stock_data['code'] == stk_codes[1]]
stock_data51['date'] = pd.to_datetime(stock_data51['date'])
stock_data51.sort_values(by=['date'], inplace=True)
stock_data51['return'] = np.log(stock_data51['close']) - np.log(stock_data51['close'].shift(periods=1))
stock_data51.dropna(inplace=True)
ind = (stock_data51['return'] >= -0.1) & (stock_data51['return'] <= 0.1)
stock_data51 = stock_data51.loc[ind, :]
stock_data51
plt.plot(stock_data51['return'].values)
Out[21]:
In [22]:
stock_data52 = stock_data[stock_data['code'] == stk_codes[2]]
stock_data52['date'] = pd.to_datetime(stock_data52['date'])
stock_data52.sort_values(by=['date'], inplace=True)
stock_data52['return'] = np.log(stock_data52['close']) - np.log(stock_data52['close'].shift(periods=1))
stock_data52.dropna(inplace=True)
ind = (stock_data52['return'] >= -0.1) & (stock_data52['return'] <= 0.1)
stock_data52 = stock_data52.loc[ind, :]
stock_data52
plt.plot(stock_data52['return'].values)
Out[22]:
In [23]:
stock_data53 = stock_data[stock_data['code'] == stk_codes[3]]
stock_data53['date'] = pd.to_datetime(stock_data53['date'])
stock_data53.sort_values(by=['date'], inplace=True)
stock_data53['return'] = np.log(stock_data53['close']) - np.log(stock_data53['close'].shift(periods=1))
stock_data53.dropna(inplace=True)
ind = (stock_data53['return'] >= -0.1) & (stock_data53['return'] <= 0.1)
stock_data53 = stock_data53.loc[ind, :]
stock_data53
plt.plot(stock_data53['return'].values)
Out[23]:
In [24]:
stock_data54 = stock_data[stock_data['code'] == stk_codes[4]]
stock_data54['date'] = pd.to_datetime(stock_data54['date'])
stock_data54.sort_values(by=['date'], inplace=True)
stock_data54['return'] = np.log(stock_data54['close']) - np.log(stock_data54['close'].shift(periods=1))
stock_data54.dropna(inplace=True)
ind = (stock_data54['return'] >= -0.1) & (stock_data54['return'] <= 0.1)
stock_data54 = stock_data54.loc[ind, :]
stock_data54
plt.plot(stock_data54['return'].values)
Out[24]:
In [26]:
stock_data50
Out[26]:
In [27]:
stock_data51
Out[27]:
In [29]:
stock_data52
Out[29]:
In [30]:
stock_data53
Out[30]:
In [31]:
stock_data54
Out[31]:
In [35]:
stock1 = pd.read_csv('1EShowData_data_stock_daily_price_2001-2018.csv',
encoding='GB2312',
usecols=[1, 5])
stock1.columns = ['date', 'close']
stock1.dropna(inplace=True)
stock1['date'] = pd.to_datetime(stock1['date'])
stock1['return'] = np.log(stock1['close']) - np.log(stock1['close'].shift(periods=1))
stock1.dropna(inplace=True)
ind = (stock1['return'] >= -0.1) & (stock1['return'] <= 0.1)
stock1 = stock1.loc[ind, :]
stock1
Out[35]:
In [41]:
index = pd.read_csv('1EShowData_data_Index_daily_price_2001-2018.csv',
encoding='GB2312',
usecols=[1, 5])
index.columns = ['date', 'close']
index['date'] = pd.to_datetime(index['date'])
index['return'] = np.log(index['close']) - np.log(index['close'].shift(periods=1))
stock1.dropna(inplace=True)
ind = (index['return'] >= -0.1) & (index['return'] <= 0.1)
index = index.loc[ind, :]
index
Out[41]:
In [38]:
stock_data
Out[38]:
In [53]:
ret_rf = stock_data[['date', 'rfreturn']]
ret_rf.dropna(inplace=True)
ret_rf.drop_duplicates(inplace=True, subset=['date'])
ret_rf.sort_values(by=['date'], inplace=True)
ret_rf['date'] = pd.to_datetime(ret_rf['date'])
ret_rf
Out[53]:
stock1, stock_data50, stock_data51, stock_data_52, stock_data_53, stock_data_53, index, rf_return
In [43]:
data_matrix = pd.merge(left=index[['date', 'return']],
right=stock1[['date', 'return']],
on='date',
how='inner',
sort=True)
data_matrix
Out[43]:
In [44]:
data_matrix = pd.merge(left=data_matrix,
right=stock_data50[['date', 'return']],
on='date',
how='inner',
sort=True)
data_matrix
Out[44]:
In [45]:
data_matrix = pd.merge(left=data_matrix,
right=stock_data51[['date', 'return']],
on='date',
how='inner',
sort=True)
data_matrix
Out[45]:
In [46]:
data_matrix = pd.merge(left=data_matrix,
right=stock_data52[['date', 'return']],
on='date',
how='inner',
sort=True)
data_matrix
Out[46]:
In [47]:
data_matrix = pd.merge(left=data_matrix,
right=stock_data53[['date', 'return']],
on='date',
how='inner',
sort=True)
data_matrix
Out[47]:
In [48]:
data_matrix = pd.merge(left=data_matrix,
right=stock_data54[['date', 'return']],
on='date',
how='inner',
sort=True)
data_matrix
Out[48]:
In [54]:
data_matrix = pd.merge(left=data_matrix,
right=ret_rf,
on='date',
how='inner',
sort=True)
data_matrix
Out[54]:
In [55]:
data_matrix.columns = ['date', 'ind', 'stk1', 'stk2', 'stk3', 'stk4', 'stk5', 'stk6', 'rf']
data_matrix
Out[55]:
In [56]:
data_matrix['ind'] = data_matrix['ind'] - data_matrix['rf']
data_matrix['stk1'] = data_matrix['stk1'] - data_matrix['rf']
data_matrix['stk2'] = data_matrix['stk2'] - data_matrix['rf']
data_matrix['stk3'] = data_matrix['stk3'] - data_matrix['rf']
data_matrix['stk4'] = data_matrix['stk4'] - data_matrix['rf']
data_matrix['stk5'] = data_matrix['stk5'] - data_matrix['rf']
data_matrix['stk6'] = data_matrix['stk6'] - data_matrix['rf']
data_matrix
Out[56]:
In [57]:
ret_ind = data_matrix['ind'].values
T = len(ret_ind)
N = 6
mu_market = np.mean(ret_ind)
sigma_market = np.sum((ret_ind - mu_market)**2)/T
ret_stocks = data_matrix[['stk1', 'stk2', 'stk3', 'stk4', 'stk5', 'stk6']].values
In [58]:
x = sm.add_constant(ret_ind)
y = ret_stocks[:, 3]
model = sm.OLS(y,x)
results = model.fit()
print(results.summary())
无限制模型参数估计
In [59]:
x = np.ones((T, 2))
x[:, 1] = ret_ind
y = ret_stocks
xTx = np.dot(np.transpose(x), x)
xTy = np.dot(np.transpose(x), y)
AB_hat = np.dot(np.linalg.inv(xTx), xTy)
ALPHA = AB_hat[0]
print(ALPHA)
BETA = AB_hat[1]
RESD = y - np.dot(x, AB_hat)
COV = np.dot(np.transpose(RESD), RESD)/T
invCOV = np.linalg.inv(COV)
限制模型参数估计
In [60]:
xr = np.ones((T, 1))
xr[:, 0] = ret_ind
yr = ret_stocks
xrTxr = np.dot(np.transpose(xr), xr)
xrTyr = np.dot(np.transpose(xr), yr)
ABr_hat = np.dot(np.linalg.inv(xrTxr), xrTyr)
RESDr = yr - np.dot(xr, ABr_hat)
COVr = np.dot(np.transpose(RESDr), RESDr)/T
invCOVr = np.linalg.inv(COVr)
In [61]:
trans_ALPHA = np.ones((len(ALPHA), 1))
trans_ALPHA[:, 0] = ALPHA
SWchi2 = T*(1/(1+mu_market**2/sigma_market))*np.dot(np.dot(ALPHA, invCOV), trans_ALPHA)
SWF = (T-N-1)/N*(1/(1+mu_market**2/sigma_market))*np.dot(np.dot(ALPHA, invCOV), trans_ALPHA)
pvalue_Wchi2 = 1 - chi2.cdf(SWchi2[0], N)
pvalue_WF = 1 - f.cdf(SWF[0], N, T-N-1)
print(pvalue_Wchi2)
print(pvalue_WF)
In [63]:
SLRchi2 = T*(np.log(np.linalg.det(COVr)) - np.log(np.linalg.det(COV)))
pvalue_SLRchi2 = 1 - chi2.cdf(SLRchi2, N)
print(pvalue_SLRchi2)
In [64]:
a = np.zeros((6, 1))
a[:, 0] = np.sum(RESDr, axis=0)
salpha = np.dot(invCOVr, a)
b = np.dot(ret_ind, RESDr)
sbeta = np.zeros((6, 1))
sbeta[:, 0] = np.dot(invCOVr, b)
score = np.concatenate((salpha, sbeta), axis=0)
In [65]:
a = np.concatenate((invCOVr*T, invCOVr*np.sum(ret_ind)), axis=1)
b = np.concatenate((invCOVr*np.sum(ret_ind), invCOVr*np.sum(ret_ind**2)), axis=1)
Minfo = np.concatenate((a, b), axis=0)
SLMchi2 = np.dot(np.dot(np.transpose(score), np.linalg.inv(Minfo)), score)
pvalue_SLMchi2 = 1 - chi2.cdf(SLMchi2[0][0], N)
print(pvalue_SLMchi2)
In [66]:
print('{:>10s}, {:>10s}, {:>10s}, {:>10s}'.format('Wald Test1', 'Wald Test2', 'LR Test', 'LM Test'))
print('{:10.5f}, {:10.5f}, {:10.5f}, {:10.5f}'.format(SWchi2[0], SWF[0], SLRchi2, SLMchi2[0][0]))
print('{:10.5f}, {:10.5f}, {:10.5f}, {:10.5f}'.format(pvalue_Wchi2, pvalue_WF, pvalue_SLRchi2, pvalue_SLMchi2))