使用statsmodels预测置信区间

Question

我正在建立一个线性模型，如下所示：

import statsmodels.api as sm
from statsmodels.stats.outliers_influence import summary_table
import numpy as np
import random

x = np.arange(1,101, 1)
y = random.sample(range(1,1000), 100)

X = sm.add_constant(x)
regr = sm.OLS(y, X)
fit = regr.fit()

st, data, ss2 = summary_table(fit, alpha=0.05)

我可以根据data确定标准误差和置信区间。

现在，我想预测新数据的置信区间，如下所示：

new_data = [102, 103, 104, 105]

fit.get_prediction(new_data)

但这返回：

Traceback (most recent call last):

  File "<ipython-input-168-372d2610946d>", line 14, in <module>
    fit.get_prediction(new)

  File "/Users/spotter/anaconda3/lib/python3.6/site-packages/statsmodels/regression/linear_model.py", line 2138, in get_prediction
    weights=weights, row_labels=row_labels, **kwds)

  File "/Users/user/anaconda3/lib/python3.6/site-packages/statsmodels/regression/_prediction.py", line 163, in get_prediction
    predicted_mean = self.model.predict(self.params, exog, **pred_kwds)

  File "/Users/user/anaconda3/lib/python3.6/site-packages/statsmodels/regression/linear_model.py", line 261, in predict
    return np.dot(exog, params)

ValueError: shapes (1,4) and (2,) not aligned: 4 (dim 1) != 2 (dim 0

Answer 1

由于使用截距训练了模型，因此在创建new_data （=添加1的列）时，还需要将其包括在内。

new_data = sm.add_constant([102, 103, 104, 105])
result = fit.get_prediction(new_data)
result.conf_int()