goodness of fit and multi-linear regression analysis

[Hui SHI]

Hi experts,

It did not work when I ran "estat gof" after a multivariable linear regression model. I did not know why and checked the textbook. I found that the "estat gof" command only was used after logistic regression model.

So I come here to ask how to test goodness of fit after linear regress model.

Thanks.

Hui

[Joseph Coveney]

rvfplot is a good start, but you might want to take a look at the others, as well.

help regress postestimation plots

[Carlo Lazzaro]

Hui:
other tests are relevant too, such as:
1) -estat hettest- (heteroskedasticity);
2) -estat vce, corr- (serial correlation of the idiosyncratic error);
3) -linktest- (specifcation of the functional form of the regressand).

[Hui SHI]

rvfplot is a good start, but you might want to take a look at the others, as well.

help regress postestimation plots

Thanks. It is very helpful.

[Hui SHI]

Hui:
other tests are relevant too, such as:
1) -estat hettest- (heteroskedasticity);
2) -estat vce, corr- (serial correlation of the idiosyncratic error);
3) -linktest- (specifcation of the functional form of the regressand).

Breusch–Pagan/Cook–Weisberg test for heteroskedasticity
Assumption: Normal error terms
Variable: Fitted values of QOL_T1

H0: Constant variance

chi2(1) = 154.17
Prob > chi2 = 0.0000

Hi Lazzaro,

I fitted three regress models. After running every model, I ran the "estat hettest", the output showed that P<0.001. Does this mean that the model is not good? What should I do next?

I see many paper used linear regression did not report the postestimation result, like heteroskedasticity test.

I have two options now as followed below.

(1) fix model (I do not know how to fix), try heteroskedasticity test.
(2) do not consider the result of heteroskedasticity test

Which one should I choose?

[Carlo Lazzaro]

Hui:
just go -robust- standard errors and do not repeat -estat hettest- anymore.