margins after suest: "Warning: cannot perform check for estimable functions"

[John Mullahy]

Can anyone point me to documentation describing why the above warning is issued when factor variables are used but not when their non-factor-variable counterparts are used? Thanks in advance for any suggestions.

Here is some sample code:

cap preserve
cap drop _all
set obs 1000
forval j=1/2 {
 gen x`j'=uniform()>.5
 gen y`j'=uniform()>.5
}
forval j=1/2 {
 qui probit y`j' x1 x2
 est store e`j'
}
qui suest e1 e2

margins, dydx(*) exp(normal(xb(e1_y1))+normal(xb(e2_y2)))

est clear

forval j=1/2 {
 qui probit y`j' i.(x1 x2)
 est store e`j'
}
qui suest e1 e2

margins, dydx(*) exp(normal(xb(e1_y1))+normal(xb(e2_y2)))

cap restore

And here are its results:

. cap preserve

. cap drop _all

. set obs 1000
number of observations (_N) was 0, now 1,000

. forval j=1/2 {
  2.  gen x`j'=uniform()>.5
  3.  gen y`j'=uniform()>.5
  4. }

. forval j=1/2 {
  2.  qui probit y`j' x1 x2
  3.  est store e`j'
  4. }

. qui suest e1 e2

.
. margins, dydx(*) exp(normal(xb(e1_y1))+normal(xb(e2_y2)))

Average marginal effects                        Number of obs     =      1,000
Model VCE    : Robust

Expression   : normal(xb(e1_y1))+normal(xb(e2_y2))
dy/dx w.r.t. : x1 x2

------------------------------------------------------------------------------
             |            Delta-method
             |      dy/dx   Std. Err.      z    P>|z|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
          x1 |   .0124886   .0449718     0.28   0.781    -.0756545    .1006318
          x2 |   .0287027   .0449262     0.64   0.523    -.0593509    .1167563
------------------------------------------------------------------------------

.
. est clear

.
. forval j=1/2 {
  2.  qui probit y`j' i.(x1 x2)
  3.  est store e`j'
  4. }

. qui suest e1 e2

.
. margins, dydx(*) exp(normal(xb(e1_y1))+normal(xb(e2_y2)))
Warning: cannot perform check for estimable functions.

Average marginal effects                        Number of obs     =      1,000
Model VCE    : Robust

Expression   : normal(xb(e1_y1))+normal(xb(e2_y2))
dy/dx w.r.t. : 1.x1 1.x2

------------------------------------------------------------------------------
             |            Delta-method
             |      dy/dx   Std. Err.      z    P>|z|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
        1.x1 |   .0124926   .0449882     0.28   0.781    -.0756827    .1006678
        1.x2 |   .0287057   .0449403     0.64   0.523    -.0593756    .1167871
------------------------------------------------------------------------------
Note: dy/dx for factor levels is the discrete change from the base level.

.
. cap restore

.
end of do-file