Extract p values in a list for Adfuller test(Test for stationarity) in ARIMA Time series modeling python pandas

问题

 Col1   Col2   Col3
  12     10     3
   3      5     2
  100    12     10

and so on.....

Code to write adfuller test for ARIMA modeling in Time series. (will calculate p value for all the columns of dataframe df)

import statsmodels.tsa.stattools as tsa
adf_results = {}
for col in df.columns.values:  
    adf_results[col] = tsa.adfuller(df[col])

Using this code I am getting outputs in below format: (output when I type adf_result)

 [IN] adf_result
 [OUT]
  {'Col1': (-4.236149193618492,
  0.0005719678593039654,  #This is the second value for this column/p value
  0,
  37,
  {'1%': -3.6209175221605827,
   '5%': -2.9435394610388332,
   '10%': -2.6104002410518627},
  138.66116123406837),
 'Col2': (-3.707023043984407,
  0.004015446231411924,  #This is the second value for this column/p value
  0,
  37,
  {'1%': -3.6209175221605827,
   '5%': -2.9435394610388332,
   '10%': -2.6104002410518627},
  144.6019873130419),
 'Col3': (1.8083888603589304,
  0.9983655107052215,   #This is the second value for this column/p value
  0,
  37,
  {'1%': -3.6209175221605827,
   '5%': -2.9435394610388332,
   '10%': -2.6104002410518627},
  -74.4384052778039)}

and so on.

In this question, second value/p value is

    0.0005719678593039654, 0.004015446231411924 and 0.9983655107052215 for the 3 columns taken.

I need columns where second value >0.05 in one list and columns where p value <0.05 in another list

So one list will be col1 and col2 (second value/p value<0.05) and the other list will be col3 (second value/p value<0.05)

回答1:

import pandas as pd
from io import StringIO


data = StringIO("""
Col1 Col2 Col3
12 10 3
3 5 2
100 12 10
13 4 1
""")

# load data into data frame
df = pd.read_csv(data, sep=' ')

import statsmodels.tsa.stattools as tsa
adf_results = {}
for col in df.columns.values:
    adf_results[col] = tsa.adfuller(df[col])

# loop over dictionary data
columns_big = []
columns_small = []
for key, value in adf_results.items():
    if value[1] > 0.05:
        columns_big.append(key)
    else:
        columns_small.append(key)

Output:

columns_big = ['Col1', 'Col3']
columns_small = ['Col2']

来源：https://stackoverflow.com/questions/60003348/extract-p-values-in-a-list-for-adfuller-testtest-for-stationarity-in-arima-tim

标签

python

pandas

machine-learning

time-series

arima