Does STATA REG provide test statistic for Fitness of model and/or homogeneity of residuals and or misspecification?

[Lucas Long]

Hello everyone,

Does STATA REG provide test statistic for Fitness of model and/or homogeneity of residuals and or misspecification?

For example, if I ran simple regression like :

reg DV IV1 IV2,

the results only show R square, is there any code that shows the Fitness or Homogeneity, or misspecification?

Thanks in adavance.

[Lucas Long]

I found that
estat imtest Can be used as to test the Homogeneity of residuals. Still have trouble to find the Fitness of Model and Misspecification.

[Lucas Long]

And

estat ovtest

can be used to test misspecification.

[Rich Goldstein]

in general, you should take a look at :

help regress postestimation
help regress postestimation plots

[Carlo Lazzaro]

Lucas:
see al also -linktest- for testing the possible misspecification of the functional form of the regressand.

[Lucas Long]

I try to use estat gof, however it won't show any results.

Anyway I can get the goodness of fit test results?

[Rich Goldstein]

please read the FAQ and show exactly what you entered and exactly what Stata replied - I don't see "estat gof" in either of the help files I sent you to so where did you find it?

[Lucas Long]

please read the FAQ and show exactly what you entered and exactly what Stata replied - I don't see "estat gof" in either of the help files I sent you to so where did you find it?

https://www.stata.com/manuals13/restatgof.pdf

[Carlo Lazzaro]

Lucas:
the first part of Rich's wise recommendation is probably the most relevant one: how could interested listers reply positively without any clues of what have you typed and what Stata returned (use CODE delimiters, please)? Thanks.

[Lucas Long]

Lucas:
the first part of Rich's wise recommendation is probably the most relevant one: how could interested listers reply positively without any clues of what have you typed and what Stata returned (use CODE delimiters, please)? Thanks.

Hi Carlo,

Thanks for your reply,
I enter :

reg dependentvariable independentvariable

estat gof 

Stata replied:

estat gof not valid
r(321);

It looks like that estat gof can only input after the

logit

function.

Thus if I run:

logit dependentvariable independentvariable

estat gof

It showes the Chi-square.

Is there any way that I can get the Chi-square by using the regular regression?

Thanks

[Carlo Lazzaro]

Lucas:
thanks for providing furthe details.
As you surmised, -esta gof- is not a post estimation commands that you can use after -regress-.
That said, if you want to check your OLS, you can use -estat hettest- (to test for heteroskedastcity) and -linktest- or -estat imtest- (to test for model misspecification).
Eventually, OLS gives back R-sq and Adj R_sq (no chi_sq).

[Lucas Long]

Lucas:
thanks for providing furthe details.
As you surmised, -esta gof- is not a post estimation commands that you can use after -regress-.
That said, if you want to check your OLS, you can use -estat hettest- (to test for heteroskedastcity) and -linktest- or -estat imtest- (to test for model misspecification).
Eventually, OLS gives back R-sq and Adj R_sq (no chi_sq).

Thank you so very much, Carlo,

One of my paper reviewers insist I put the Chi-square for model fitness, that's why I am trying to ask.

Thanks again,

[Rich Goldstein]

ask the reviewer for a citation which you can then share with us; as Carlo Lazzaro says, this command cannot be used after -regress- (and I recommend against it anyway as it is heavily affected by the number of "groups" used; I repeat that you should see the help files I noted in #2 above (as well as

help linktest

as noted by Carlo

it may also help if you can provide the context the reviewer is responding to

[Carlo Lazzaro]

Lucas:

If you actually performed an OLS and you have provided the list with all the relevant details concerning the matters, it sounds really strange that a decent reviewer asks for something that does not make sense.
If that were the case, as an aside to @Rich Goldstein 's helpful advice, you can rebut the reviewer in a gentle way that, while, OLS resturns R-sq, some post estimation tests to be conducted after OLS actually return a chi2 statistic, as you can see from the following toy example:

. sysuse auto.dta
(1978 Automobile Data)

. regress price i.foreign

      Source |       SS           df       MS      Number of obs   =        74
-------------+----------------------------------   F(1, 72)        =      0.17
       Model |  1507382.66         1  1507382.66   Prob > F        =    0.6802
    Residual |   633558013        72  8799416.85   R-squared       =    0.0024
-------------+----------------------------------   Adj R-squared   =   -0.0115
       Total |   635065396        73  8699525.97   Root MSE        =    2966.4

------------------------------------------------------------------------------
       price |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
     foreign |
    Foreign  |   312.2587   754.4488     0.41   0.680    -1191.708    1816.225
       _cons |   6072.423    411.363    14.76   0.000     5252.386     6892.46
------------------------------------------------------------------------------

. estat hettest

Breusch-Pagan / Cook-Weisberg test for heteroskedasticity
         Ho: Constant variance
         Variables: fitted values of price

         chi2(1)      =     0.85
         Prob > chi2  =   0.3554

. estat imtest

Cameron & Trivedi's decomposition of IM-test

---------------------------------------------------
              Source |       chi2     df      p
---------------------+-----------------------------
  Heteroskedasticity |       0.43      1    0.5103
            Skewness |      13.02      1    0.0003
            Kurtosis |       4.20      1    0.0404
---------------------+-----------------------------
               Total |      17.66      3    0.0005
---------------------------------------------------

.

See also https://stats.stackexchange.com/ques...and-chi-square

[Lucas Long]

Lucas:

If you actually performed an OLS and you have provided the list with all the relevant details concerning the matters, it sounds really strange that a decent reviewer asks for something that does not make sense.
If that were the case, as an aside to @Rich Goldstein 's helpful advice, you can rebut the reviewer in a gentle way that, while, OLS resturns R-sq, some post estimation tests to be conducted after OLS actually return a chi2 statistic, as you can see from the following toy example:

. sysuse auto.dta
(1978 Automobile Data)

. regress price i.foreign

Source | SS df MS Number of obs = 74
-------------+---------------------------------- F(1, 72) = 0.17
Model | 1507382.66 1 1507382.66 Prob > F = 0.6802
Residual | 633558013 72 8799416.85 R-squared = 0.0024
-------------+---------------------------------- Adj R-squared = -0.0115
Total | 635065396 73 8699525.97 Root MSE = 2966.4

------------------------------------------------------------------------------
price | Coef. Std. Err. t P>|t| [95% Conf. Interval]
-------------+----------------------------------------------------------------
foreign |
Foreign | 312.2587 754.4488 0.41 0.680 -1191.708 1816.225
_cons | 6072.423 411.363 14.76 0.000 5252.386 6892.46
------------------------------------------------------------------------------

. estat hettest

Breusch-Pagan / Cook-Weisberg test for heteroskedasticity
Ho: Constant variance
Variables: fitted values of price

chi2(1) = 0.85
Prob > chi2 = 0.3554

. estat imtest

Cameron & Trivedi's decomposition of IM-test

---------------------------------------------------
Source | chi2 df p
---------------------+-----------------------------
Heteroskedasticity | 0.43 1 0.5103
Skewness | 13.02 1 0.0003
Kurtosis | 4.20 1 0.0404
---------------------+-----------------------------
Total | 17.66 3 0.0005
---------------------------------------------------

.

See also https://stats.stackexchange.com/ques...and-chi-square

This is great and is exactly what I was looking for!

Thank you so very much Carlo!