Conformability error with margins after repost with buildfvinfo

[Alex Ballantyne]

Hi,

I am writing a bespoke estimation that uses clogit as a base. After running clogit I modify the e(b) and e(V) matrices and use ereturn repost . If I include the option buildfvinfo when I repost the matrices, the margins postestimation command no longer works and returns "conformability error; r(503);".
I have checked this in a very straightforward setting using clogit and ologit, where I do not modify the b and V matrices but simply copy and repost them. Without the buildfvinfo option it works fine, but with it margins breaks. Any advice would be much appreciated.

Cheers,
-Alex

//Test eret repost do file
clear all
set more off

use http://www.stata-press.com/data/r13/clogitid

clogit y x1 x2, group(id)
margins, dydx(*) predict(pu0)

eretswap

margins, dydx(*) predict(pu0)

end

// Program to copy and repost e(b) and e(V)
program eretswap, eclass
    syntax [, *]

mat tempb = e(b)
mat tempV = e(V)

eret repost b=tempb V=tempV, buildfvinfo

end

[Alex Ballantyne]

Some further investigations have found that the internal matrix H changes dimension after the ereturn repost command. If the H matrix after the initial estimation is dimension MxM, after reposting it is (M-1)x(M-1). In addition, if I use factor variable notation on x1 (i.e. clogit y i.x1 x, group(id) ), not only does the dimension of H change, but the contents also contains elements that are not -1,0,1.

Hopefully this additional information is useful in isolating the problem.

[Joerg Luedicke (StataCorp)]

To make this work, Alex can add the undocumented ADDCONS (sic) option to ereturn repost. Here is a modified version of his code

. // Program to copy and repost e(b) and e(V)
. program eretswap, eclass
  1.         syntax [, *]
  2.         mat tempb = e(b)
  3.         mat tempV = e(V)
  4.         eret repost b=tempb V=tempV, buildfvinfo ADDCONS
  5. end

. 
. // Program to copy and repost e(b) and e(V)
. use http://www.stata-press.com/data/r13/clogitid

. clogit y x1 x2, group(id)
note: multiple positive outcomes within groups encountered.

Iteration 0:   log likelihood = -123.42828  
Iteration 1:   log likelihood = -123.41386  
Iteration 2:   log likelihood = -123.41386  

Conditional (fixed-effects) logistic regression

                                                Number of obs     =        369
                                                LR chi2(2)        =       9.07
                                                Prob > chi2       =     0.0107
Log likelihood = -123.41386                     Pseudo R2         =     0.0355

------------------------------------------------------------------------------
           y |      Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
          x1 |    .653363   .2875215     2.27   0.023     .0898312    1.216895
          x2 |   .0659169   .0449555     1.47   0.143    -.0221943    .1540281
------------------------------------------------------------------------------

. margins, dydx(*) predict(pu0)

Average marginal effects                        Number of obs     =        369
Model VCE    : OIM

Expression   : Pr(y|fixed effect is 0), predict(pu0)
dy/dx w.r.t. : x1 x2

------------------------------------------------------------------------------
             |            Delta-method
             |      dy/dx   Std. Err.      z    P>|z|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
          x1 |   .1465753   .0625489     2.34   0.019     .0239817     .269169
          x2 |   .0147878   .0090277     1.64   0.101    -.0029061    .0324817
------------------------------------------------------------------------------

. eretswap

. margins, dydx(*) predict(pu0)

Average marginal effects                        Number of obs     =        369
Model VCE    : OIM

Expression   : Pr(y|fixed effect is 0), predict(pu0)
dy/dx w.r.t. : x1 x2

------------------------------------------------------------------------------
             |            Delta-method
             |      dy/dx   Std. Err.      z    P>|z|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
          x1 |   .1465753   .0625489     2.34   0.019     .0239817     .269169
          x2 |   .0147878   .0090277     1.64   0.101    -.0029061    .0324817
------------------------------------------------------------------------------

Joerg