if prob>F=0,07, the model is ok?

[Jessie Cai]

i already run a hausman test ,and the result is use fixed model. i also want use time-fixed model. but the prob>F in time-fixed model is 0,07. is this ok?

[Carlo Lazzaro]

Jessie:
what if you omit -i.year-?

[Jessie Cai]

Jessie:
what if you omit -i.year-?

this is my code and result. and i need use this time-fixed effect model to draw marginsplot.

[Carlo Lazzaro]

Jessie:
I can only add that the F-test shows that you cannot be confident that your coefficients, when taken jointly, differ from zero.
This is just an information, neither good, nor bad..

[Jessie Cai]

Jessie:
I can only add that the F-test shows that you cannot be confident that your coefficients, when taken jointly, differ from zero.
This is just an information, neither good, nor bad..

do i need the model prob>F always < 0,05, or if prob>F <0,1, the model can also be used?

[Carlo Lazzaro]

Jessie:
the problem is not about using or not the model, but what the F-test outcome tells you.
in your case, your model can be safely replaced by - xtreg gini, fe- (ie, a sort of mean for panel data).

[Jessie Cai]

Jessie:
the problem is not about using or not the model, but what the F-test outcome tells you.
in your case, your model can be safely replaced by - xtreg gini, fe- (ie, a sort of mean for panel data).

thank you for your help! i have another question, so i compare 4 models, this means fe model is the best one?

[Carlo Lazzaro]

Jessie:
it depends.
In panel data analysis, pooled OLS is usually outperformed by -xtreg-.
Under -xtreg-, -fe-and -re- specification are not interchangeable.
You do not provide any detail about postestimation: for instance, have you investigated non-linearity in your regression models?