Effects option does not take effects for predict method?

[FeiYao-Edinburgh]

Hi there,

I am using linearmodels to modelling panel data. I just found that the predict method returned me exactly identical results whether I set effects option True or False. They are identical to fitted_values, too, if predictions were made on model data. I believe they are all predictions made by slope and common constant only? Below I provide my codes and relevant outputs for you to check. If no difference, then I do not comprehend the significance of effects option. Or does this option has some other meanings that I do not know?

• Codes

from linearmodels.datasets import wage_panel
import pandas as pd
data = wage_panel.load()
data = data.set_index(["nr", "year"])

from linearmodels.panel import PanelOLS
exog_vars = ["expersq", "union", "married"]
exog = sm.add_constant(data[exog_vars])
mod = PanelOLS(data.lwage, exog, entity_effects=True)
fe_res = mod.fit()

fe_res.predict(exog.head(), effects = False)
fe_res.predict(exog.head(), effects = True)
fe_res.fitted_values.head()

• Outputs
  

Thanks in advance!

[bashtage]

Effects are errors and so are not part of the prediction. They are not consistently estimated in the classic fixed T, large N setting. If one dimension is fixed and the other is large, so that the fixed dimension can be consistently estimated, then include these are dummy variables (e.g., time dummies in a large N setting.

[bashtage]

But yes, this seems suspicion.

[FeiYao-Edinburgh]

Effects are errors and so are not part of the prediction.

Thanks for your clarification!

[bashtage]

I need to check what this is supposed to do; at a minimum the docstring is inadequate.

[bashtage]

Thanks for the report!

[bashtage]

And the reproducible example.