通常在Barra 模型中，因子暴露直接来自基本面或者技术面数据本身，而在学术界通常使用时序回归得到因子暴露，在进行截面回归得到因子收益率。
截面回归
1.先对时序时间求个股收益均值
In [13]:
start_date='2018-01-10'
end_date='2019-01-10'
start_date=datetime.datetime.strptime(start_date,'%Y-%m-%d')
end_date=datetime.datetime.strptime(end_date,'%Y-%m-%d')+datetime.timedelta(days=1)
stock_list2=list(stock_params.index)
price = get_price(stock_list2, start_date,end_date, '1d', ['quote_rate'], True, None, is_panel=1)['quote_rate']
return_df=price.apply(lambda x:x.mean(),axis=0)
return_df.head()
Out[13]:
000001.SZ -0.072040
000002.SZ -0.091996
000063.SZ -0.196550
000069.SZ -0.129263
000100.SZ -0.125031
dtype: float64
In [14]:
from statsmodels import regression
import statsmodels.api as sm
stock_params['return_ave_stock']=return_df
x=np.array(stock_params['coef'])
x = x.astype(float)
y=np.array(stock_params['return_ave_stock'])
y = y.astype(float)
#X = sm.add_constant(x)
est=sm.OLS(y,x)
est=est.fit()
est.summary()
Out[14]:
| Dep. Variable: | y | R-squared: | 0.114 |
|---|---|---|---|
| Model: | OLS | Adj. R-squared: | 0.109 |
| Method: | Least Squares | F-statistic: | 27.30 |
| Date: | Wed, 20 Feb 2019 | Prob (F-statistic): | 4.15e-07 |
| Time: | 09:58:20 | Log-Likelihood: | 74.090 |
| No. Observations: | 214 | AIC: | -146.2 |
| Df Residuals: | 213 | BIC: | -142.8 |
| Df Model: | 1 | ||
| Covariance Type: | nonrobust |
| | coef | std err | t | P>|t| | [0.025 | 0.975] | | -- | ------- | ------- | ------ | ------- | ------ | ------ | | x1 | -0.0874 | 0.017 | -5.224 | 0.000 | -0.120 | -0.054 |
| Omnibus: | 141.322 | Durbin-Watson: | 1.507 |
|---|---|---|---|
| Prob(Omnibus): | 0.000 | Jarque-Bera (JB): | 2215.624 |
| Skew: | -2.231 | Prob(JB): | 0.00 |
| Kurtosis: | 18.118 | Cond. No. | 1.00 |
因子收益率
In [15]:
second_params2 = pd.DataFrame(columns = ["coef"],index=['factor'])
x=stock_params['coef'].reshape(-1,1)
y=stock_params['return_ave_stock'].reshape(-1,1)
model = LinearRegression(fit_intercept=None)
model.fit(x,y)
second_params2.ix[0,'coef']=model.coef_[0][0]
second_params2
Out[15]:
| coef | |
|---|---|
| factor | -0.0874333 |
2.残差的概率密度分布
In [16]:
pd.Series(est.resid).plot.density(figsize = (20, 8))
plt.title('resid_density',size=25)
#fig = plt.figure(figsize = (20, 8))
#ax = est.resid.plot.kde(label = 'Beta')
#ax.set_title('Beta',size=25)
#ax.legend()
Out[16]:
<matplotlib.text.Text at 0x7fa6bc2d3630>
3. 对残差进行t检验
In [17]:
from scipy import stats
stat_val, p_value = stats.ttest_1samp(est.resid, 0)
p_value
Out[17]:
1.5905611338293164e-14
In [18]:
stat_val
Out[18]:
-8.254519611155741
In [19]:
ttest_df1= pd.DataFrame(columns = ["stats_val","p_value"],index=['factor_residual'])
ttest_df1.ix[0,"stats_val"]=stat_val
ttest_df1.ix[0,"p_value"]=p_value
ttest_df1
Out[19]:
| stats_val | p_value | |
|---|---|---|
| factor_residual | -8.25452 | 1.59056e-14 |
截面回归下三因子模型
1.第一步，时序回归
In [20]:
flag1=True
while start_date<=end_date:
next_date=start_date+datetime.timedelta(days=1)
#print(next_date)
#取全市场股票各股票市值因子
df = get_fundamentals(query(asharevalue.symbol, asharevalue.total_mv).filter(\
asharevalue.symbol.in_(stock_list1)).order_by(asharevalue.total_mv.asc()),date=start_date)
df.set_index('asharevalue_symbol',inplace=True)
df.index.name='symbol'
df2=get_fundamentals(query(asharevalue.symbol,balance.total_equity/asharevalue.total_mv).filter(\
asharevalue.symbol.in_(stock_list1)),date=start_date)
df2.set_index('asharevalue_symbol',inplace=True)
df2.index.name='symbol'
df2.columns=['BM']
df2.sort_values(by=['BM'],inplace=True)
stocklist_cap_top=list(df[:int(len(df)*0.3)].index)
stocklist_cap_bottom=list(df[int(len(df)*0.7):].index)
stocklist_BM_bottom=list(df2[:int(len(df)*0.3)].index)
stocklist_BM_top=list(df2[int(len(df)*0.7):].index)
try:
value_top = get_price(stocklist_cap_top,None, start_date, '1d', \
['quote_rate'],True, None,1,is_panel=1)['quote_rate']
return_sum_top=value_top.apply(lambda x:x.mean(),axis=1)
value_bottom = get_price(stocklist_cap_bottom,None, start_date, '1d', \
['quote_rate'], True, None,1,is_panel=1)['quote_rate']
return_sum_bottom=value_bottom.apply(lambda x:x.mean(),axis=1)
return_factor=return_sum_top-return_sum_bottom
value_top_BM = get_price(stocklist_BM_top,None, start_date, '1d', \
['quote_rate'],True, None,1,is_panel=1)['quote_rate']
return_sum_top_BM=value_top_BM.apply(lambda x:x.mean(),axis=1)
value_bottom_BM = get_price(stocklist_BM_bottom,None, start_date, '1d', \
['quote_rate'], True, None,1,is_panel=1)['quote_rate']
return_sum_bottom_BM=value_bottom_BM.apply(lambda x:x.mean(),axis=1)
return_factor_BM=return_sum_top_BM-return_sum_bottom_BM
except:
start_date=start_date+datetime.timedelta(days=1)
continue
if flag1:
flag1=False
return_date_total_factor = return_factor
return_date_total_factor_BM = return_factor_BM
else:
return_date_total_factor = return_date_total_factor.append(return_factor)
return_date_total_factor_BM = return_date_total_factor_BM.append(return_factor_BM)
start_date=start_date+datetime.timedelta(days=1)
start_date_market='2018-01-10'
end_date_market='2019-01-10'
start_date=datetime.datetime.strptime(start_date_market,'%Y-%m-%d')
end_date=datetime.datetime.strptime(end_date_market,'%Y-%m-%d')+datetime.timedelta(days=1)
return_market= get_price('000300.SH', start_date_market,end_date_market, '1d', ['quote_rate'], True, None, is_panel=1)['quote_rate']
return_date_total_factor_market=pd.DataFrame(return_market)
return_date_total_factor=pd.DataFrame(return_date_total_factor)
return_date_total_factor_BM=pd.DataFrame(return_date_total_factor_BM)
In [21]:
return_date_total_factor_df= pd.concat([return_date_total_factor_market,return_date_total_factor,return_date_total_factor_BM],axis=1)
return_date_total_factor_df.columns=['market_factor_return','cap_factor_return','BM_factor_return']
return_date_total_factor_df.fillna(0,inplace=True)
return_date_total_factor_df.head()
Out[21]:
| market_factor_return | cap_factor_return | BM_factor_return | |
|---|---|---|---|
| 2018-01-10 | 0.4420 | -1.201322 | 0.040815 |
| 2018-01-11 | -0.0529 | 0.445029 | -0.470802 |
| 2018-01-12 | 0.4616 | -0.906181 | -0.052597 |
| 2018-01-15 | 0.0056 | -2.711759 | 0.527817 |
| 2018-01-16 | 0.7866 | -0.314921 | 1.009860 |
In [22]:
stock_params = pd.DataFrame(columns = ["intercept", "coef_market","coef_cap","coef_BM"],index=stock_list)
i=0
for stock in price.columns:
x=return_date_total_factor_df.ix[:,('market_factor_return','cap_factor_return','BM_factor_return')]
y=price[stock].reshape(-1,1)
model = LinearRegression()
try:
model.fit(x,y)
stock_params.ix[i,'intercept'] = model.intercept_[0]
stock_params.ix[i,'coef_market'] = model.coef_[0][0]
stock_params.ix[i,'coef_cap'] = model.coef_[0][1]
stock_params.ix[i,'coef_BM'] = model.coef_[0][2]
except:
stock_params.ix[i,'intercept'] =np.nan
stock_params.ix[i,'coef_market'] =np.nan
stock_params.ix[i,'coef_cap'] = np.nan
stock_params.ix[i,'coef_BM'] = np.nan
i=i+1
stock_params.dropna(inplace=True)
stock_params.head()
Out[22]:
| intercept | coef_market | coef_cap | coef_BM | |
|---|---|---|---|---|
| 601088.SH | 0.0642632 | 1.31645 | -0.422633 | 0.640151 |
| 002624.SZ | 0.0635011 | 1.5009 | -0.430519 | 0.612894 |
| 600583.SH | 0.0498274 | 1.25657 | 0.0752462 | 0.415214 |
| 600383.SH | 0.0842546 | 1.02056 | 0.654638 | 0.151889 |
| 601633.SH | 0.0737299 | 0.731708 | 0.406155 | 0.151578 |
In [23]:
plt.figure(figsize=(20,16))
plt.subplot(311)
stock_params['coef_market'].plot.kde(label = 'Beta_market',color='g')
plt.title('Beta_market',size=25)
plt.subplot(312)
stock_params['coef_cap'].plot.kde(label = 'Beta_cap',color='r')
plt.title('Beta_cap',size=25)
plt.subplot(313)
stock_params['coef_BM'].plot.kde(label = 'Beta_BM')
plt.title('Beta_BM',size=25)
Out[23]:
<matplotlib.text.Text at 0x7fa6b8478d30>
2.第二步，截面回归
In [24]:
stock_params['return_ave_stock']=return_df
stock_params.dropna(inplace=True)
x=stock_params.ix[:,('coef_market','coef_cap','coef_BM')]
x = x.astype(float)
y=np.array(stock_params['return_ave_stock'])
y = y.astype(float)
#X = sm.add_constant(x)
est=sm.OLS(y,x)
est=est.fit()
est.summary()
Out[24]:
| Dep. Variable: | y | R-squared: | 0.419 |
|---|---|---|---|
| Model: | OLS | Adj. R-squared: | 0.408 |
| Method: | Least Squares | F-statistic: | 37.56 |
| Date: | Wed, 20 Feb 2019 | Prob (F-statistic): | 2.51e-18 |
| Time: | 10:01:41 | Log-Likelihood: | 82.155 |
| No. Observations: | 159 | AIC: | -158.3 |
| Df Residuals: | 156 | BIC: | -149.1 |
| Df Model: | 3 | ||
| Covariance Type: | nonrobust |
| | coef | std err | t | P>|t| | [0.025 | 0.975] | | ----------- | ------- | ------- | ------ | ------- | ------ | ------ | | coef_market | -0.1139 | 0.013 | -8.889 | 0.000 | -0.139 | -0.089 | | coef_cap | -0.0192 | 0.021 | -0.902 | 0.368 | -0.061 | 0.023 | | coef_BM | 0.0332 | 0.022 | 1.511 | 0.133 | -0.010 | 0.077 |
| Omnibus: | 137.536 | Durbin-Watson: | 2.075 |
|---|---|---|---|
| Prob(Omnibus): | 0.000 | Jarque-Bera (JB): | 2463.668 |
| Skew: | -2.982 | Prob(JB): | 0.00 |
| Kurtosis: | 21.339 | Cond. No. | 2.15 |
因子收益率
In [25]:
second_params3 = pd.DataFrame(columns = ["coef_market","coef_cap","coef_BM"],index=['factor_return'])
x=stock_params.ix[:,('coef_market','coef_cap','coef_BM')]
y=stock_params['return_ave_stock'].reshape(-1,1)
model = LinearRegression(fit_intercept=None)
model.fit(x,y)
second_params3.ix[0,'coef_market']=model.coef_[0][0]
second_params3.ix[0,'coef_cap']=model.coef_[0][1]
second_params3.ix[0,'coef_BM']=model.coef_[0][2]
second_params3
Out[25]:
| coef_market | coef_cap | coef_BM | |
|---|---|---|---|
| factor_return | -0.113863 | -0.019238 | 0.0332294 |
In [26]:
pd.Series(est.resid).plot.density(figsize = (20, 8))
plt.title('resid_density_3factor',size=25)
Out[26]:
<matplotlib.text.Text at 0x7fa6b83849e8>
In [28]:
stat_val, p_value = stats.ttest_1samp(est.resid, 0)
ttest_df2= pd.DataFrame(columns = ["stats_val","p_value"],index=['factor_residual'])
ttest_df2.ix[0,"stats_val"]=stat_val
ttest_df2.ix[0,"p_value"]=p_value
ttest_df2
Out[28]:
| stats_val | p_value | |
|---|---|---|
| factor_residual | -0.234637 | 0.814794 |
截面回归下五因子模型
In [36]:
flag1=True
while start_date<=end_date:
next_date=start_date+datetime.timedelta(days=1)
#print(next_date)
#取全市场股票各股票市值因子
df = get_fundamentals(query(asharevalue.symbol, asharevalue.total_mv).filter(\
asharevalue.symbol.in_(stock_list1)).order_by(asharevalue.total_mv.asc()),date=start_date)
df.set_index('asharevalue_symbol',inplace=True)
df.index.name='symbol'
df2=get_fundamentals(query(asharevalue.symbol,balance.total_equity/asharevalue.total_mv).filter(\
asharevalue.symbol.in_(stock_list1)),date=start_date)
df2.set_index('asharevalue_symbol',inplace=True)
df2.index.name='symbol'
df2.columns=['BM']
df2.sort_values(by=['BM'],inplace=True)
df3 = get_fundamentals(query(asharevalue.symbol, ashareprofit.roe_ttm).filter(\
asharevalue.symbol.in_(stock_list1)).order_by(ashareprofit.roe_ttm.desc()),date=start_date)
df3.set_index('asharevalue_symbol',inplace=True)
df3.index.name='symbol'
df4 = get_fundamentals(query(asharevalue.symbol, asharevalue.pe_ttm).filter(\
asharevalue.symbol.in_(stock_list1)).order_by(asharevalue.pe_ttm.asc()),date=start_date)
df4.set_index('asharevalue_symbol',inplace=True)
df4.index.name='symbol'
stocklist_cap_top=list(df[:int(len(df)*0.3)].index)
stocklist_cap_bottom=list(df[int(len(df)*0.7):].index)
stocklist_BM_bottom=list(df2[:int(len(df)*0.3)].index)
stocklist_BM_top=list(df2[int(len(df)*0.7):].index)
stocklist_roe_top=list(df3[:int(len(df)*0.3)].index)
stocklist_roe_bottom=list(df3[int(len(df)*0.7):].index)
stocklist_invest_top=list(df4[:int(len(df)*0.3)].index)
stocklist_invest_bottom=list(df4[int(len(df)*0.7):].index)
try:
#cap因子收益
value_top = get_price(stocklist_cap_top,None, start_date, '1d', \
['quote_rate'],True, None,1,is_panel=1)['quote_rate']
return_sum_top=value_top.apply(lambda x:x.mean(),axis=1)
value_bottom = get_price(stocklist_cap_bottom,None, start_date, '1d', \
['quote_rate'], True, None,1,is_panel=1)['quote_rate']
return_sum_bottom=value_bottom.apply(lambda x:x.mean(),axis=1)
return_factor=return_sum_top-return_sum_bottom
#BM因子收益
value_top_BM = get_price(stocklist_BM_top,None, start_date, '1d', \
['quote_rate'],True, None,1,is_panel=1)['quote_rate']
return_sum_top_BM=value_top_BM.apply(lambda x:x.mean(),axis=1)
value_bottom_BM = get_price(stocklist_BM_bottom,None, start_date, '1d', \
['quote_rate'], True, None,1,is_panel=1)['quote_rate']
return_sum_bottom_BM=value_bottom_BM.apply(lambda x:x.mean(),axis=1)
return_factor_BM=return_sum_top_BM-return_sum_bottom_BM
#roe因子收益
value_top_roe = get_price(stocklist_roe_top,None, start_date, '1d', \
['quote_rate'],True, None,1,is_panel=1)['quote_rate']
return_sum_top_roe=value_top_roe.apply(lambda x:x.mean(),axis=1)
value_bottom_roe = get_price(stocklist_roe_bottom,None, start_date, '1d', \
['quote_rate'], True, None,1,is_panel=1)['quote_rate']
return_sum_bottom_roe=value_bottom_roe.apply(lambda x:x.mean(),axis=1)
return_factor_roe=return_sum_top_roe-return_sum_bottom_roe
#投资效率因子
value_top_invest = get_price(stocklist_invest_top,None, start_date, '1d', \
['quote_rate'],True, None,1,is_panel=1)['quote_rate']
return_sum_top_invest=value_top_invest.apply(lambda x:x.mean(),axis=1)
value_bottom_invest = get_price(stocklist_invest_bottom,None, start_date, '1d', \
['quote_rate'], True, None,1,is_panel=1)['quote_rate']
return_sum_bottom_invest=value_bottom_invest.apply(lambda x:x.mean(),axis=1)
return_factor_invest=return_sum_top_invest-return_sum_bottom_invest
except:
start_date=start_date+datetime.timedelta(days=1)
continue
if flag1:
flag1=False
return_date_total_factor = return_factor
return_date_total_factor_BM = return_factor_BM
return_date_total_factor_roe = return_factor_roe
return_date_total_factor_invest = return_factor_invest
else:
return_date_total_factor = return_date_total_factor.append(return_factor)
return_date_total_factor_BM = return_date_total_factor_BM.append(return_factor_BM)
return_date_total_factor_roe = return_date_total_factor_roe.append(return_factor_roe)
return_date_total_factor_invest = return_date_total_factor_invest.append(return_factor_invest)
start_date=start_date+datetime.timedelta(days=1)
start_date_market='2018-01-10'
end_date_market='2019-01-10'
start_date=datetime.datetime.strptime(start_date_market,'%Y-%m-%d')
end_date=datetime.datetime.strptime(end_date_market,'%Y-%m-%d')+datetime.timedelta(days=1)
return_market= get_price('000300.SH', start_date_market,end_date_market, '1d', ['quote_rate'], True, None, is_panel=1)['quote_rate']
return_date_total_factor_market=pd.DataFrame(return_market)
return_date_total_factor=pd.DataFrame(return_date_total_factor)
return_date_total_factor_BM=pd.DataFrame(return_date_total_factor_BM)
return_date_total_factor_roe=pd.DataFrame(return_date_total_factor_roe)
return_date_total_factor_invest=pd.DataFrame(return_date_total_factor_invest)
In [37]:
return_date_total_factor_df= pd.concat([return_date_total_factor_market,return_date_total_factor,return_date_total_factor_BM,return_date_total_factor_roe,return_date_total_factor_invest],axis=1)
return_date_total_factor_df.columns=['market_factor_return','cap_factor_return','BM_factor_return','roe_factor_return','invest_factor_return']
return_date_total_factor_df.fillna(0,inplace=True)
return_date_total_factor_df.head()
Out[37]:
| market_factor_return | cap_factor_return | BM_factor_return | roe_factor_return | invest_factor_return | |
|---|---|---|---|---|---|
| 2018-01-10 | 0.4420 | -1.201322 | 0.040815 | 0.510692 | 0.761227 |
| 2018-01-11 | -0.0529 | 0.445029 | -0.470802 | -0.749758 | -0.708931 |
| 2018-01-12 | 0.4616 | -0.906181 | -0.052597 | 0.713779 | 0.727008 |
| 2018-01-15 | 0.0056 | -2.711759 | 0.527817 | 1.180434 | 1.751605 |
| 2018-01-16 | 0.7866 | -0.314921 | 1.009860 | -0.723669 | 0.532430 |
In [38]:
stock_params = pd.DataFrame(columns = ["intercept", "coef_market","coef_cap","coef_BM","coef_roe","coef_invest"],index=stock_list)
i=0
for stock in price.columns:
x=return_date_total_factor_df.ix[:,('market_factor_return','cap_factor_return','BM_factor_return','roe_factor_return','invest_factor_return')]
y=price[stock].reshape(-1,1)
model = LinearRegression()
try:
model.fit(x,y)
stock_params.ix[i,'intercept'] = model.intercept_[0]
stock_params.ix[i,'coef_market'] = model.coef_[0][0]
stock_params.ix[i,'coef_cap'] = model.coef_[0][1]
stock_params.ix[i,'coef_BM'] = model.coef_[0][2]
stock_params.ix[i,'coef_roe'] = model.coef_[0][3]
stock_params.ix[i,'coef_invest'] = model.coef_[0][4]
except:
stock_params.ix[i,'intercept'] =np.nan
stock_params.ix[i,'coef_market'] =np.nan
stock_params.ix[i,'coef_cap'] = np.nan
stock_params.ix[i,'coef_BM'] = np.nan
stock_params.ix[i,'coef_roe'] = np.nan
stock_params.ix[i,'coef_invest'] = np.nan
i=i+1
stock_params.dropna(inplace=True)
stock_params.head()
Out[38]:
| intercept | coef_market | coef_cap | coef_BM | coef_roe | coef_invest | |
|---|---|---|---|---|---|---|
| 601088.SH | 0.0557343 | 1.31861 | -0.4687 | 0.635007 | -0.0442248 | -0.032073 |
| 002624.SZ | 0.162543 | 1.4749 | 0.120437 | 0.741134 | 0.58148 | 0.310008 |
| 600583.SH | 0.173552 | 1.21508 | 0.923075 | 1.25812 | 1.40324 | -0.23489 |
| 600383.SH | 0.0730138 | 1.02584 | 0.550943 | -0.0387876 | -0.240604 | 0.125318 |
| 601633.SH | 0.0670419 | 0.727927 | 0.46702 | 0.562513 | 0.376727 | -0.403318 |
In [39]:
stock_params['return_ave_stock']=return_df
stock_params.dropna(inplace=True)
x=stock_params.ix[:,('coef_market','coef_cap','coef_BM','coef_roe','coef_invest')]
x = x.astype(float)
y=np.array(stock_params['return_ave_stock'])
y = y.astype(float)
#X = sm.add_constant(x)
est=sm.OLS(y,x)
est=est.fit()
est.summary()
Out[39]:
| Dep. Variable: | y | R-squared: | 0.420 |
|---|---|---|---|
| Model: | OLS | Adj. R-squared: | 0.401 |
| Method: | Least Squares | F-statistic: | 22.29 |
| Date: | Wed, 20 Feb 2019 | Prob (F-statistic): | 9.04e-17 |
| Time: | 10:29:01 | Log-Likelihood: | 82.217 |
| No. Observations: | 159 | AIC: | -154.4 |
| Df Residuals: | 154 | BIC: | -139.1 |
| Df Model: | 5 | ||
| Covariance Type: | nonrobust |
| | coef | std err | t | P>|t| | [0.025 | 0.975] | | ----------- | ------- | ------- | ------ | ------- | ------ | ------ | | coef_market | -0.1141 | 0.013 | -8.496 | 0.000 | -0.141 | -0.088 | | coef_cap | -0.0219 | 0.025 | -0.862 | 0.390 | -0.072 | 0.028 | | coef_BM | 0.0348 | 0.023 | 1.504 | 0.135 | -0.011 | 0.080 | | coef_roe | -0.0095 | 0.022 | -0.436 | 0.663 | -0.053 | 0.034 | | coef_invest | 0.0194 | 0.023 | 0.861 | 0.391 | -0.025 | 0.064 |
| Omnibus: | 137.364 | Durbin-Watson: | 2.067 |
|---|---|---|---|
| Prob(Omnibus): | 0.000 | Jarque-Bera (JB): | 2452.053 |
| Skew: | -2.977 | Prob(JB): | 0.00 |
| Kurtosis: | 21.294 | Cond. No. | 2.90 |
In [40]:
second_params3 = pd.DataFrame(columns = ["coef_market","coef_cap","coef_BM","coef_roe","coef_invest"],index=['factor_return'])
x=stock_params.ix[:,('coef_market','coef_cap','coef_BM',"coef_roe","coef_invest")]
y=stock_params['return_ave_stock'].reshape(-1,1)
model = LinearRegression(fit_intercept=None)
model.fit(x,y)
second_params3.ix[0,'coef_market']=model.coef_[0][0]
second_params3.ix[0,'coef_cap']=model.coef_[0][1]
second_params3.ix[0,'coef_BM']=model.coef_[0][2]
second_params3.ix[0,'coef_roe']=model.coef_[0][3]
second_params3.ix[0,'coef_invest']=model.coef_[0][4]
second_params3
Out[40]:
| coef_market | coef_cap | coef_BM | coef_roe | coef_invest | |
|---|---|---|---|---|---|
| factor_return | -0.114142 | -0.021943 | 0.0347699 | -0.00950722 | 0.019449 |
In [ ]:
查看以上策略详情请到supermind量化交易官网查看：多因子专题（一）-- 截面回归 附源代码
