我正在尝试运行多变量回归并得到错误:
“ ValueError:endog和exog矩阵的大小不同”
我的代码段如下:
df_raw = pd.DataFrame(data=df_raw)
y = (df_raw['daily pct return']).astype(float)
x1 = (df_raw['Excess daily return']).astype(float)
x2 = (df_raw['Excess weekly return']).astype(float)
x3 = (df_raw['Excess monthly return']).astype(float)
x4 = (df_raw['Trading vol / mkt cap']).astype(float)
x5 = (df_raw['Std dev']).astype(float)
x6 = (df_raw['Residual risk']).astype(float)
y = y.replace([np.inf, -np.inf],np.nan).dropna()
print(y.shape)
print(x1.shape)
print(x2.shape)
print(x3.shape)
print(x4.shape)
print(x5.shape)
print(x6.shape)
df_raw.to_csv('Raw_final.csv', header=True)
result = smf.OLS(exog=y, endog=[x1, x2, x3, x4, x5, x6]).fit()
print(result.params)
print(result.summary())
从我的代码中可以看到,我正在检查每个变量的“形状”。我得到以下输出,该错误指示原因是y变量只有48392个值,而所有其他变量都有48393个值:
(48392,) (48393,) (48393,) (48393,) (48393,) (48393,) (48393,)
我的数据框如下所示:
daily pct return | Excess daily return | weekly pct return | index weekly pct return | Excess weekly return | monthly pct return | index monthly pct return | Excess monthly return | Trading vol / mkt cap | Std dev
------------------|---------------------|-------------------|-------------------------|----------------------|--------------------|--------------------------|-----------------------|-----------------------|-------------
| | | | | | | | 0.207582827 |
0.262658228 | 0.322397801 | | | | | | | 0.285585677 |
0.072681704 | 0.126445534 | | | | | | | 0.272920624 |
0.135514019 | 0.068778682 | | | | | | | 0.213149083 |
-0.115226337 | -0.173681889 | | | | | | | 0.155653699 |
-0.165116279 | -0.176569405 | | | | | | | 0.033925024 |
0.125348189 | 0.079889239 | | | | | | | 0.030968484 | 0.544133212
0.022277228 | -0.044949678 | | | | | | | 0.020735381 | 0.385659608
0.150121065 | 0.102119782 | | | | | | | 0.063563881 | 0.430868447
0.336842105 | 0.333590483 | | | | | | | 0.210193049 | 0.893734807
0.011023622 | -0.011860658 | 0.320987654 | -0.657089012 | 0.978076666 | | | | 0.100468109 | 1.137976483
0.37694704 | 0.308505907 | | | | | | | 0.135828281 | 1.867394416
有人能解决矩阵大小的问题吗,这样我就不再收到此错误了?我想我需要从y变量(“每日pct返回”)中删除值APART的第一行,但是我不确定如何实现此目标?
提前谢谢!
答案 0 :(得分:0)
我假设您想丢弃与y值无穷大相关的所有数据。
df_raw = pd.DataFrame(data=df_raw)
df_raw['daily pct return']) = df_raw['daily pct return']).astype(float).replace([np.inf, -np.inf],np.nan)
df_raw = df_raw.dropna()
然后继续进行回归。
答案 1 :(得分:0)
终于解决了这个问题!有三个问题:
1)y变量的大小为48392,而其他6个变量的大小均为48393。为解决此问题,我添加了以下代码行以删除第一行:
var toyota = {
make: "Toyota",
model: "Corolla",
fuel: 0,
tank: function(addingfuel) {
this.fuel = this.fuel + parseInt(addingfuel);
},
start: function() {
if (this.fuel === 0) {
alert("stop");
} else {
alert("go");
}
},
};
var addingfuel = prompt("Please enter fuel added", "liter");
toyota.tank(addingfuel); // you need to pass this otherwise it is undefined
toyota.start();
2)我的数据框有很多空单元格。除非每个单元格都有一个值,否则您无法执行回归。因此,我提供了一些代码,以用NaN替换所有infs和空单元格,然后用0值填充所有NaN。代码段:
df_raw = df_raw.drop([0])
3)我编写多元回归公式的方式是错误的。我将其纠正如下:
df_raw ['daily pct return']= df_raw ['daily pct return'].replace([np.inf, -np.inf],np.nan)
df_raw = df_raw.replace(r'\s+', np.nan, regex=True).replace('', np.nan)
df_raw.fillna(value=0, axis=1,inplace=True)
总而言之,我的更新代码如下:
result = smf.ols(formula='y ~ x1 + x2 + x3 + x4 + x5 + x6', data=df_raw).fit()