Hello, I am trying to have the partial effect of being an immigrant evaluated at the mean level of ysm (years since migration. I have pooled immigrants and natives in one regression. For natives, ysm has value of zero.

Code:
. svy: regress lnhourlyw c.educ c.exper c.exper#c.exper i.ib2.em02 i.ib2.bgu1 i.b
> 017 i.immigrant i.immigrant#c.ysm i.immigrant#c.ysm#c.ysm i.year
(running regress on estimation sample)

Survey: Linear regression

Number of strata   =         1                 Number of obs     =     120,875
Number of PSUs     =   120,875                 Population size   =  14,718,062
                                               Design df         =     120,874
                                               F(  21, 120854)   =     3530.20
                                               Prob > F          =      0.0000
                                               R-squared         =      0.4754

--------------------------------------------------------------------------------
               |             Linearized
     lnhourlyw |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
---------------+----------------------------------------------------------------
          educ |   .0632044   .0003949   160.05   0.000     .0624304    .0639785
         exper |   .0357629   .0004329    82.62   0.000     .0349145    .0366113
               |
       c.exper#|
       c.exper |  -.0005102   8.63e-06   -59.09   0.000    -.0005271   -.0004933
               |
          em02 |
         1-19  |  -.0746067    .002862   -26.07   0.000    -.0802162   -.0689972
  100 or more  |    .076627   .0025719    29.79   0.000     .0715862    .0816678
               |
          bgu1 |
no supervis..  |  -.0894801   .0023106   -38.73   0.000    -.0940088   -.0849515
executive m..  |     .14258   .0035159    40.55   0.000     .1356889    .1494712
               |
          b017 |
      higheco  |   .0615848   .0025705    23.96   0.000     .0565467     .066623
               |
     immigrant |
      foreign  |  -.0837934   .0051853   -16.16   0.000    -.0939564   -.0736303
               |
     immigrant#|
         c.ysm |
       native  |          0  (omitted)
      foreign  |  -.0106163   .0006566   -16.17   0.000    -.0119031   -.0093295
               |
     immigrant#|
   c.ysm#c.ysm |
       native  |          0  (omitted)
      foreign  |   .0002425   .0000167    14.49   0.000     .0002098    .0002753
               |
          year |
         2005  |   .0085027   .0050942     1.67   0.095    -.0014819    .0184873
         2006  |   .0089234   .0053721     1.66   0.097    -.0016059    .0194527
         2007  |   .0227152   .0054473     4.17   0.000     .0120385    .0333919
         2008  |   .0357659   .0055571     6.44   0.000      .024874    .0466579
         2009  |   .0445801   .0054383     8.20   0.000     .0339211     .055239
         2010  |   .0404299   .0047779     8.46   0.000     .0310654    .0497944
         2011  |   .0501135   .0046862    10.69   0.000     .0409287    .0592984
         2012  |   .0564489    .004543    12.43   0.000     .0475447    .0653531
         2013  |   .0598773   .0045868    13.05   0.000     .0508872    .0688674
         2014  |   .0624139   .0047666    13.09   0.000     .0530714    .0717564
               |
         _cons |   2.471768   .0075186   328.75   0.000     2.457031    2.486504
-------------------------------------------------------------------------------
Now, I would like to have the partial effect as mentioned in the beginning. However, it is not estimable.

Code:
.  margins, dydx(immigrant) at(ysm=(10.96))

Average marginal effects                        Number of obs     =    120,875
Model VCE    : Linearized

Expression   : Linear prediction, predict()
dy/dx w.r.t. : 1.immigrant
at           : ysm             =       10.96

------------------------------------------------------------------------------
             |            Delta-method
             |      dy/dx   Std. Err.      t    P>|t|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
   immigrant |
    foreign  |          .  (not estimable)
------------------------------------------------------------------------------
Note: dy/dx for factor levels is the discrete change from the base level.
I don't know, why it is not estimable. Any clue?