Announcement

Collapse
No announcement yet.
X
  • Filter
  • Time
  • Show
Clear All
new posts

  • No F-Test value

    Hello,

    I've performed a linear regression analyses, that reaches an R-squared of 0.22. But the field for the F-value (18, 124) is left empty in the stata output. Can anyone tell me, why this is the case?

    Thanks!

  • #2
    Showing your commands and output could greatly help. This might or might not apply to your case.

    http://www.stata.com/statalist/archi.../msg00646.html
    -------------------------------------------
    Richard Williams, Notre Dame Dept of Sociology
    Stata Version: 17.0 MP (2 processor)

    EMAIL: [email protected]
    WWW: https://www3.nd.edu/~rwilliam

    Comment


    • #3
      Hi,

      I've attached the output and command as photo. Thanks for your hint. I do have included one dummy variable, that is 1 in only one single case. However, removing that variable from the model does not help to get a F-value

      Comment


      • #4
        Christopher:
        my gut-feeling is that you have too many clusters (50) for such a limited number of observations (159).
        Does F-value come alive when default SEs are invoked?
        Kind regards,
        Carlo
        (Stata 18.0 SE)

        Comment


        • #5
          Thanks Carlo, indeed you are right. When I do not cluster SEs, then the F-value gets calculated. Hmm. So what can I do? Of course I want to cluster and in my eyes 50 clusters for 159 observations is not so uncommon, since panel data with a mean of three observed years is not that seldomn...

          Comment


          • #6
            Christopher:
            the issue with clustered SEs may rest on a negligible between-cluster variance.
            Has any variable repeated values within the 3-year span of time?
            Did you perform an acid test about this via -xtsum-?
            Does anything changes when you switch to -xtreg-?
            Kind regards,
            Carlo
            (Stata 18.0 SE)

            Comment


            • #7
              FWIW, I think the ancient Statalist post by Vince that Richard linked to in #2 is the explanation. For some reason (could be singleton dummies, could be something else), the G matrix in Vince's post that is the filling of the robust VCE sandwich DGD is not full rank. This isn't necessarily a problem for tests of individual parameters for the reasons Vince suggested, but Christopher wants an F stat for the model. It looks like that's not going to be possible, because an F stat for the model means a joint test of all the regressors, and that can't be done because #regressors > rank(G).

              One last thought - in the regress output, the coeff on xtc7 is essentially zero. Looks like it could be roundoff error rather than a real coefficient estimate. Maybe if you drop this coefficient (should probably do that anyway), the F stat will also reappear.

              Comment


              • #8
                Thanks for your comments.
                I found out that I can still calculate the F-Value using the "test" command. Could that be a propper way to solve the problem?
                http://www.stata.com/statalist/archi.../msg00583.html

                Comment


                • #9
                  Christopher:
                  you can safely follow the road pointed out by Roger, as in the following example:
                  Code:
                  . use auto.dta, clear
                  (1978 Automobile Data)
                  
                  . reg price foreign mpg
                  
                        Source |       SS       df       MS              Number of obs =      74
                  -------------+------------------------------           F(  2,    71) =   14.07
                         Model |   180261702     2  90130850.8           Prob > F      =  0.0000
                      Residual |   454803695    71  6405685.84           R-squared     =  0.2838
                  -------------+------------------------------           Adj R-squared =  0.2637
                         Total |   635065396    73  8699525.97           Root MSE      =  2530.9
                  
                  ------------------------------------------------------------------------------
                         price |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
                  -------------+----------------------------------------------------------------
                       foreign |   1767.292    700.158     2.52   0.014     371.2169    3163.368
                           mpg |  -294.1955   55.69172    -5.28   0.000    -405.2417   -183.1494
                         _cons |   11905.42   1158.634    10.28   0.000     9595.164    14215.67
                  ------------------------------------------------------------------------------
                  
                  . testparm mpg foreign
                  
                   ( 1)  foreign = 0
                   ( 2)  mpg = 0
                  
                         F(  2,    71) =   14.07
                              Prob > F =    0.0000
                  or, sticking with your original query:
                  Code:
                  . use auto.dta, clear
                  (1978 Automobile Data)
                  
                  . reg price foreign mpg, cluster(foreign)
                  
                  Linear regression                                      Number of obs =      74
                                                                         F(  0,     1) =       .
                                                                         Prob > F      =       .
                                                                         R-squared     =  0.2838
                                                                         Root MSE      =  2530.9
                  
                                                  (Std. Err. adjusted for 2 clusters in foreign)
                  ------------------------------------------------------------------------------
                               |               Robust
                         price |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
                  -------------+----------------------------------------------------------------
                       foreign |   1767.292   195.3672     9.05   0.070     -715.084    4249.668
                           mpg |  -294.1955   39.50161    -7.45   0.085    -796.1111    207.7201
                         _cons |   11905.42   783.1955    15.20   0.042     1953.973    21856.86
                  ------------------------------------------------------------------------------
                  
                  . testparm mpg foreign
                  
                   ( 1)  foreign = 0
                   ( 2)  mpg = 0
                         Constraint 2 dropped
                  
                         F(  1,     1) =   81.83
                              Prob > F =    0.0701
                  Last edited by Carlo Lazzaro; 06 Mar 2015, 09:43.
                  Kind regards,
                  Carlo
                  (Stata 18.0 SE)

                  Comment


                  • #10
                    I do have included one dummy variable, that is 1 in only one single case. However, removing that variable from the model does not help to get a F-value
                    FWIW, if there are any singleton clusters, that, too, will cause G to be singular. The solution then is to exclude those clusters or, if possible, combine them with other reasonably similar clusters.

                    Comment


                    • #11
                      In Carlo's output above (the last one): Why is constraint 2 dropped? Is it possibly the existence of singleton clusters?
                      Can the calculated F-value (1, 1) be used for describing the model that should have an f-value (0, 1)?

                      Comment


                      • #12
                        Christopher:
                        as far as your first question is concerned, constraint #2 is dropped because, as reported in -help j_robustsingular-:
                        Are you using a svy estimator or did you specify the vce(cluster clustvar) option?

                        The VCE you have just estimated is not of sufficient rank to perform the model test. As discussed in [R] test, the model test with clustered or survey data is distributed as
                        F(k,d-k+1) or chi2(k), where k is the number of constraints and d=number of clusters or d=number of PSUs minus the number of strata. Because the rank of the VCE is at most d and
                        the model test reserves 1 degree of freedom for the constant, at most d-1 constraints can be tested, so k must be less than d. The model that you just fit does not meet this
                        requirement.
                        Sorry, but I cannot follow your second question.
                        Kind regards,
                        Carlo
                        (Stata 18.0 SE)

                        Comment


                        • #13
                          Thank you!

                          My second question was: Can the F-Value of 81.83 be used in order to describe the model "reg price foreign mpg, cluster(foreign)", that reports the F-value as missing? I guess it is a good approximation.

                          Comment


                          • #14
                            Christopher:
                            I would be cautious, because the toy model is basically ill-specified.
                            At the top of that, even with one single constraint te F-test is not statistical significant (favouring an intercept-only model).
                            In conclusion, considering a limping F-test as an acceptable proxy of what should have been in a ideal scenario is, at the end of the day, a matter of personal judgement call.
                            Kind regards,
                            Carlo
                            (Stata 18.0 SE)

                            Comment


                            • #15
                              Thank you!

                              Comment

                              Working...
                              X