Question

I am using the pandas.ols function from version 0.7.3. I seem to be getting inconsistent values for adjusted $R^2$ when using the simple regression vs. window regression. For instance, if realizedData and pastData have 600 entries then

model = pandas.ols(y = realizedData, x = pastData, intercept = 0, window = 600)

produces the following output:-

-------------------------Summary of Regression Analysis-------------------------

Formula: Y ~ <1> + <10> + <90000>

Number of Observations:         596
Number of Degrees of Freedom:   3

R-squared:         0.6914
Adj R-squared:     0.6904

Rmse:            699.4880

F-stat (3, 593):   664.3691, p-value:     0.0000

Degrees of Freedom: model 2, resid 593

-----------------------Summary of Estimated Coefficients------------------------
      Variable       Coef    Std Err     t-stat    p-value    CI 2.5%   CI 97.5%
--------------------------------------------------------------------------------
             1     0.4171     0.0428       9.75     0.0000     0.3333     0.5010
            10     0.4362     0.0688       6.34     0.0000     0.3014     0.5709
         90000     0.0623     0.0319       1.95     0.0517    -0.0003     0.1249
---------------------------------End of Summary---------------------------------

while just using

model = pandas.ols(y = realizedData, x = pastData, intercept = 0)

gives:-

-------------------------Summary of Regression Analysis-------------------------

Formula: Y ~ <1> + <10> + <90000>

Number of Observations:         596
Number of Degrees of Freedom:   3

R-squared:         0.6914
Adj R-squared:     0.3053

Rmse:            699.4880

F-stat (3, 593):     1.7909, p-value:     0.1477

Degrees of Freedom: model 2, resid 593

-----------------------Summary of Estimated Coefficients------------------------
      Variable       Coef    Std Err     t-stat    p-value    CI 2.5%   CI 97.5%
--------------------------------------------------------------------------------
             1     0.4171     0.0428       9.75     0.0000     0.3333     0.5010
            10     0.4362     0.0688       6.34     0.0000     0.3014     0.5709
         90000     0.0623     0.0319       1.95     0.0517    -0.0003     0.1249
---------------------------------End of Summary---------------------------------

Note that the output is identical except for the adjsuted $R^2$ value.

Is this a bug or am I doing something wrong?

Answer 1

I think it has to do with the lack of intercept. Can you please report an issue on GitHub?