Error message - 'equation [kappa] not found r(303) - when choosing parametric model in Stata

[Charlotte Owen]

Hello all,

I am having trouble with a command in Stata and was wondering if anybody could shed some light on it. I am fitting a nested accelerated failure time model and have successfully managed to fit it, however, for this method I would like to test which parametric model is most appropriate e.g. generalized gamma, Weibull, Lognormal etc and I am following the advice in Cleves et al. (2010) An introduction to survival analysis, the similar advice can also be found here: https://jdemeritt.weebly.com/uploads...parametric.pdf (page 7).

My issue is that when I try to test whether k=1 using this code test [kappa]_b[_cons]=1 I get the error message 'equation [kappa] not found
r(303);'. I'm not sure why this is happening, I think maybe due to mis-specification somewhere but I'm unable to find where. Here is the model output for the generalized gamma model.

Generalized gamma AFT regression

No. of subjects = 2,379 Number of obs = 5,123
No. of failures = 2,745
Time at risk = 4435723
Wald chi2(24) = 266.91
Log pseudolikelihood = -4388.0882 Prob > chi2 = 0.0000

(Std. Err. adjusted for 2,379 clusters in ID)
-----------------------------------------------------------------------------------------
| Robust
_t | Coef. Std. Err. z P>|z| [95% Conf. Interval]
------------------------+----------------------------------------------------------------
sex |
M | .0975589 .0895883 1.09 0.276 -.0780309 .2731487
|
clinic |
KC | -.4006913 .113282 -3.54 0.000 -.6227199 -.1786626
KS | -.184139 .1198321 -1.54 0.124 -.4190056 .0507275
|
otype |
FO | .611605 .1783377 3.43 0.001 .2620695 .9611404
KAFO | -.0402873 .0963903 -0.42 0.676 -.229209 .1486343
Shoe Raise | -.9145288 .1695116 -5.40 0.000 -1.246765 -.5822922
SFAB | -.5030514 .1934337 -2.60 0.009 -.8821744 -.1239284
Spinal | .5159427 .1986001 2.60 0.009 .1266936 .9051917
Other | .1200314 .1477773 0.81 0.417 -.1696068 .4096697
|
agecat |
18-29 | .3240066 .1171543 2.77 0.006 .0943883 .5536248
30-44 | .5876781 .1697069 3.46 0.001 .2550587 .9202975
45+ | .8975508 .1624355 5.53 0.000 .579183 1.215919
|
diag |
Trauma/injury | -.7675417 .548409 -1.40 0.162 -1.842404 .3073201
Other congenital | -.1317484 .3092856 -0.43 0.670 -.737937 .4744401
Cerebral Palsy | -.0415777 .2206258 -0.19 0.851 -.4739964 .390841
Club foot | -.8421487 .2640103 -3.19 0.001 -1.359599 -.3246979
Paralysis | .0372566 .2872067 0.13 0.897 -.5256582 .6001714
Dislocation/fracture | -.6898822 .2753318 -2.51 0.012 -1.229523 -.1502418
Polio | -.7046541 .2605685 -2.70 0.007 -1.215359 -.1939492
Scoliosis/curved spine | -.5017567 .3061357 -1.64 0.101 -1.101772 .0982582
Short leg | .3748963 .3047463 1.23 0.219 -.2223955 .9721881
Stroke | .4969035 .4470183 1.11 0.266 -.3792363 1.373043
Other | .1593808 .2461399 0.65 0.517 -.3230445 .6418062
Missing | .0372867 .3296631 0.11 0.910 -.6088411 .6834145
|
_cons | 7.376842 .235195 31.36 0.000 6.915868 7.837816
------------------------+----------------------------------------------------------------
/lnsigma | .1670233 .0301981 5.53 0.000 .107836 .2262105
/kappa | .760035 .0838385 9.07 0.000 .5957146 .9243554
------------------------+----------------------------------------------------------------
sigma | 1.181782 .0356876 1.113865 1.25384
-----------------------------------------------------------------------------------------

So, please could someone advise why this might be the case? I have also tried this code test _b[kappa:_cons]=1 but I receive the same error message.

Many thanks,
Charlotte

[Andrew Musau]

You are not showing your command line which makes it difficult to help you. In general, you want the option -coeflegend- to identify how your coefficients are named.

webuse cancer, clear
streg drug age, distribution(ggamma) coeflegend
test _b[/kappa]=1

Res.:

. streg drug age, distribution(ggamma) coeflegend

         failure _d:  died
   analysis time _t:  studytime

Fitting constant-only model:

Iteration 0:   log likelihood = -61.342985  
Iteration 1:   log likelihood = -60.523452  
Iteration 2:   log likelihood = -60.489447  
Iteration 3:   log likelihood = -60.489405  
Iteration 4:   log likelihood = -60.489405  

Fitting full model:

Iteration 0:   log likelihood = -60.489405  (not concave)
Iteration 1:   log likelihood = -49.141476  
Iteration 2:   log likelihood = -43.458207  
Iteration 3:   log likelihood = -42.649202  
Iteration 4:   log likelihood = -42.619672  
Iteration 5:   log likelihood = -42.619647  
Iteration 6:   log likelihood = -42.619647  

Generalized gamma AFT regression

No. of subjects =           48                  Number of obs    =          48
No. of failures =           31
Time at risk    =          744
                                                LR chi2(2)       =       35.74
Log likelihood  =   -42.619647                  Prob > chi2      =      0.0000

------------------------------------------------------------------------------
          _t |      Coef.  Legend
-------------+----------------------------------------------------------------
        drug |   .7861776  _b[drug]
         age |  -.0644245  _b[age]
       _cons |   5.108207  _b[_cons]
-------------+----------------------------------------------------------------
    /lnsigma |  -.5174028  _b[/lnsigma]
      /kappa |   .8532814  _b[/kappa]
-------------+----------------------------------------------------------------
       sigma |   .5960666
------------------------------------------------------------------------------

. 
. test _b[/kappa]=1

 ( 1)  [/]kappa = 1

           chi2(  1) =    0.09
         Prob > chi2 =    0.7651

.

[Charlotte Owen]

Hi Andrew,

Thanks for your response. Adding the -coeflegend- option to find the exact name of the coefficients has helped!

Charlotte