Getting Durbin-Watson figure from statsmodels.api

Question

I can't extract the durbin-watson as a value on it's own from the statsmodel.api, or find anywhere any documentation to help (i found alot of documentation on it's parent library, but i couldn't decode any of it).

The value is being calculated and can be seen by doing the following model summary (i've been following the guidance here: https://www.statology.org/durbin-watson-test-python/)

from statsmodels.formula.api import ols

#fit multiple linear regression model
model = ols('rating ~ points + assists + rebounds', data=df).fit()

#view model summary
print(model.summary())

however i just can't pull out that one figure. Any ideas?

Also - it looks like DW works on a confidence interval. Can you 'set' the model to work on say 95% confidence? I essentially want to perform the test multiple times and if the DW figure is in the 95% CI, return a yes or no to continue the program.

Thanks

Kevin S · Accepted Answer · 2022-03-10 09:32:08Z

2

You need to use the durbin_watson function directly.

from statsmodels.formula.api import ols
from statsmodels.stats.stattools import durbin_watson
import pandas as pd
import numpy as np

df = pd.DataFrame(np.random.standard_normal((500,4)))
df.columns = ["rating", "points", "assists", "rebounds"]


#fit multiple linear regression model
model = ols('rating ~ points + assists + rebounds', data=df)
res = model.fit()

dw = durbin_watson(res.resid)
print(f"Durbin-Watson: {dw}")

which produces

Durbin-Watson: 1.9818102986170278

Critical values for DW statistics are not available in statsmodels. The Ljung-Box test for serial correlation is a more general approach that has critical values available.

from statsmodels.stats.diagnostic import acorr_ljungbox
lb = acorr_ljungbox(res.resid)
print(lb)

which gives

     lb_stat  lb_pvalue
1   0.003400   0.953500
2   0.774305   0.678988
3   1.412020   0.702720
4   1.890551   0.755881
5   2.176684   0.824197
6   2.397583   0.879749
7   3.186928   0.867188
8   3.639602   0.888089
9   3.793818   0.924451
10  5.639786   0.844565

The left column is the test statistic for no serial correlation and the right is the p-value.

answered Mar 10, 2022 at 9:32

Kevin S

2,82320 silver badges24 bronze badges

Sign up to request clarification or add additional context in comments.

1 Comment

GlassShark1 Over a year ago

exactly what I need - thanks :)

Collectives™ on Stack Overflow

Getting Durbin-Watson figure from statsmodels.api

1 Answer 1

1 Comment

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

1 Comment

Your Answer

Sign up or log in

Post as a guest

Related