单因子测试（下）——回归测试法

  1def MultiPeriodFactorTest(factor,DateStart,DateEnd,if_reciprocal,if_neutral_industry=True, if_neutral_mktcap=True):
  2
  3
  4    # 获取交易日序列
  5    BargainDate = w.tdays(DateStart, DateEnd, "Period=M")
  6    BargainDate = pd.DataFrame(BargainDate.Data[0],columns = ['date'])
  7    # 循环调用单期因子测试函数，得到收益率序列，IC序列，t值序列
  8    result = pd.DataFrame(columns=["DateStart","DateEnd","factor_return","t_values","IC"])
  9    for i in range(1,BargainDate.shape[0]):
 10        datebuy = BargainDate.date[i-1]
 11        datesell = BargainDate.date[i]
 12        result1 = SingelePeriodFactorTest(factor,datebuy,datesell,if_neutral_industry, if_neutral_mktcap,if_reciprocal)
 13        result = result.append(result1)
 14
 15    result.factor_return = result.factor_return
 16
 17    # 计算均值
 18    t_mean = result.t_values.mean()      
 19    return_mean =  result.factor_return.mean()   
 20    t_abs_mean = result.t_values.abs().mean()
 21
 22
 23    # t>0比例
 24    if_t_0 = pd.DataFrame.mean(result.t_values>0)
 25
 26
 27#     IC统计量
 28    IC_mean = result.IC.mean()
 29    IC_std = result.IC.std()
 30    if_IC_0 = pd.DataFrame.mean(result.IC>0)
 31    if_abs_IC_002 = pd.DataFrame.mean(result.IC.abs()>0.02)
 32
 33
 34    # 计算ICIR
 35    ICIR = IC_mean/IC_std
 36
 37    final = {"因子收益序列t均值":t_mean,
 38             "因子收益序列均值":return_mean,
 39             "t>0比例":if_t_0,
 40             "abs(t)均值":t_abs_mean,
 41             "IC均值":IC_mean,
 42             "IC标准差":IC_std,
 43             "IC>0比例":if_IC_0,
 44             "abs(IC)>0.02比例":if_abs_IC_002,
 45             "ICIR":ICIR
 46             }
 47#    '''
 48#    作图
 49#    '''
 50#    X = np.arange(result.shape[0])
 51#    
 52#    # 因子收益
 53#    plt.figure()
 54#    plt.subplot(221)
 55#    plt.bar(X,result.factor_return)
 56#    plt.title('factor_return')
 57#    plt.subplot(222)
 58#    plt.hist(result.factor_return)
 59#    plt.title('factor_return')
 60#    
 61#    # t值
 62#    plt.subplot(223)
 63#    plt.bar(X,abs(result.t_values))
 64#    plt.title('abs(t_value)')
 65#    
 66#    
 67#    # IC
 68#    plt.subplot(224)
 69#    plt.bar(X,result.IC)
 70#    plt.title('IC')
 71#    plt.tight_layout(pad=0.4, w_pad=0.5, h_pad=1.0)
 72#    
 73
 74    return(final,result)
 75
 76'''
 77 单期因子测试函数
 78'''
 79def SingelePeriodFactorTest(factor,DateStart,DateEnd,if_neutral_industry, if_neutral_mktcap,if_reciprocal):
 80
 81    factor_name = factor['value'][0][0]
 82
 83    print(factor_name + '单因子测试：' + str(DateStart) + ' -- ' + str(DateEnd))
 84    data = getData(factor,DateStart)
 85    # 是否取倒数
 86    if if_reciprocal:
 87        data.iloc[:,0] = 1/data.iloc[:,0] 
 88    # 通过本地数据库计算股票收益率
 89    stock_list = data.index.tolist()
 90    stock_list = list(map(lambda x:x.strip('.SZ'),stock_list))
 91    stock_list = list(map(lambda x:x.strip('.SH'),stock_list))
 92    stock_list = tuple(stock_list)
 93    price_s = pd.read_sql("select stockcode,closeprice as price_s \
 94                           from database\
 95                           where stockcode in {} and bargaindate='{}'\
 96                           order by stockcode".format((stock_list),DateStart),stkbase)
 97    price_s=price_s.set_index('stockcode')
 98    price_e = pd.read_sql("select stockcode,closeprice as price_e \
 99                           from database\
100                           where stockcode in {} and bargaindate='{}'\
101                           order by stockcode".format((stock_list),DateEnd),stkbase)
102
103
104
105    price_e=price_e.set_index('stockcode')
106    price=pd.merge(price_s,price_e,left_index=True,right_index=True)
107
108
109    price['rt']=price['price_e']/price['price_s']-1
110    # 标准化
111    factor_all= norm(data,if_neutral_industry, if_neutral_mktcap)  
112    factor_all.index.name = 'stockcode'
113    factor_all = factor_all.reset_index()
114    factor_all['stockcode'] = factor_all['stockcode'].apply(lambda x:x.strip('.SZ|.SH'))
115    factor_all = factor_all.set_index('stockcode')
116    Alldata = pd.merge(factor_all,price,left_index=True,right_index=True,how = 'left')
117
118   
119   
120
121    y = Alldata[['rt']]
122    X = factor_all
123    X['Intercept'] = 1
124
125    model = sml.OLS(y,X)
126    result=model.fit()
127
128
129    # 计算IC,IR
130    IC = calc_IC(factor_all[factor_name], y, 'Spearman')
131
132    print(str(DateStart) + "——" +str(DateEnd)+","+factor_name + '因子收益为:'+ str(round(result.params[0],4)*100) +'%')
133
134    return_t=pd.DataFrame([[DateStart,DateEnd,result.params[0],result.tvalues[0],IC]],
135                           columns=["DateStart","DateEnd","factor_return","t_values","IC"])
136
137    return(return_t)