为什么因果影响包不承认我的数据中的日期索引？

Question

我遇到了因果影响包无法识别数据中日期索引的问题。

我得到输入数据索引中不存在的错误20210626。下面的blob包括错误的跟踪。

​

ValueError                                Traceback (most recent call last)
Input In [97], in <cell line: 3>()
      1 pre_period = ['20210626','20210628']
      2 post_period = ['20210629','20210702']
----> 3 ci = CausalImpact(data, pre_period, post_period)

File ~/homebrew/lib/python3.8/site-packages/causalimpact/main.py:228, in CausalImpact.__init__(self, data, pre_period, post_period, model, alpha, **kwargs)
    227 def __init__(self, data, pre_period, post_period, model=None, alpha=0.05, **kwargs):
--> 228     checked_input = self._process_input_data(
    229         data, pre_period, post_period, model, alpha, **kwargs
    230     )
    231     super(CausalImpact, self).__init__(**checked_input)
    232     self.model_args = checked_input['model_args']

File ~/homebrew/lib/python3.8/site-packages/causalimpact/main.py:377, in CausalImpact._process_input_data(self, data, pre_period, post_period, model, alpha, **kwargs)
    374     raise ValueError('{args} input cannot be empty'.format(
    375                      args=', '.join(none_args)))
    376 processed_data = self._format_input_data(data)
--> 377 pre_data, post_data = self._process_pre_post_data(processed_data, pre_period,
    378                                                   post_period)
    379 alpha = self._process_alpha(alpha)
    380 model_args = self._process_model_args(**kwargs)

File ~/homebrew/lib/python3.8/site-packages/causalimpact/main.py:658, in CausalImpact._process_pre_post_data(self, data, pre_period, post_period)
    637 def _process_pre_post_data(self, data, pre_period, post_period):
    638     """
    639     Checks `pre_period`, `post_period` and returns data sliced accordingly to  each
    640     period.
   (...)
    656       ValueError: if pre_period last value is bigger than post intervention period.
    657     """
--> 658     checked_pre_period = self._process_period(pre_period, data)
    659     checked_post_period = self._process_period(post_period, data)
    661     if checked_pre_period[1] > checked_post_period[0]:

File ~/homebrew/lib/python3.8/site-packages/causalimpact/main.py:727, in CausalImpact._process_period(self, period, data)
    725         if isinstance(point, pd.Timestamp):
    726             point = point.strftime('%Y%m%d')
--> 727         raise ValueError("{point} not present in input data index.".format(
    728             point=str(point)
    729             )
    730         )
    731 if isinstance(period[0], str) or isinstance(period[0], pd.Timestamp):
    732     period = self._convert_str_period_to_int(period, data)

ValueError: 20210626 not present in input data index.

​

代码和示例数据如下所示。有人能帮忙吗？

import numpy as np
import pandas as pd
import tensorflow as tf
import tensorflow_probability as tfp
import matplotlib.pyplot as plt
from causalimpact import CausalImpact

data = pd.read_csv('~/datasets/results_covariates.csv',encoding='utf-8')
data.set_index('DT', inplace=True, drop=False)

pre_period = ['20210626','20210628']
post_period = ['20210629','20210702']
ci = CausalImpact(data, pre_period, post_period)

DT  Y   X1  X2
6/26/21 1016.15 8616.033333 164
6/27/21 1174.983333 18156.85    444
6/28/21 56571.43333 417270.6    11664
6/29/21 64821.75    420466.3167 11322
6/30/21 178269.8    2331084.75  66434
7/1/21  62314.28333 391890.9    11221
7/2/21  141387.3833 1286635.85  35207

Answer 1

设法让这件事奏效了。

因果关系影响期望索引为int、str或pd.Timestamp，在实际代码中读取csv /数据库连接器后，dt列必须重新格式化。

data['DT'] = pd.to_datetime(data['DT'])