Hi,

I am using SEM to estimate an autoregressive model:

Code:
sem (r9cesdnmv2 <- relwmo2v2c) (r10cesdnmv2 <-  r9cesdnmv2) ///
    (r11cesdnmv2 <- r10cesdnmv2) (r12cesdnmv2 <- r11cesdnmv2) ///
    (r13cesdnmv2 <- r12cesdnmv2) (r14cesdnmv2 <- r13cesdnmv2)
Here is the result:

Code:
Endogenous variables

Observed:  r9cesdnmv2 r10cesdnmv2 r11cesdnmv2 r12cesdnmv2 r13cesdnmv2 r14cesdnmv2

Exogenous variables

Observed:  relwmo2v2c

Fitting target model:

Iteration 0:   log likelihood = -36612.044  
Iteration 1:   log likelihood = -36612.044  

Structural equation model                       Number of obs     =      3,124
Estimation method  = ml
Log likelihood     = -36612.044

-----------------------------------------------------------------------------------
                  |                 OIM
                  |      Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]
------------------+----------------------------------------------------------------
Structural        |
  r9cesdnmv2      |
       relwmo2v2c |  -.4159013   .0773182    -5.38   0.000    -.5674421   -.2643604
            _cons |    1.51768   .0680087    22.32   0.000     1.384386    1.650975
  ----------------+----------------------------------------------------------------
  r10cesdnmv2     |
       r9cesdnmv2 |    .520521   .0147684    35.25   0.000     .4915755    .5494665
            _cons |   .5202732   .0321207    16.20   0.000     .4573178    .5832286
  ----------------+----------------------------------------------------------------
  r11cesdnmv2     |
      r10cesdnmv2 |   .5672095   .0147685    38.41   0.000     .5382637    .5961553
            _cons |   .5198662     .03115    16.69   0.000     .4588134    .5809191
  ----------------+----------------------------------------------------------------
  r12cesdnmv2     |
      r11cesdnmv2 |   .5626509   .0152267    36.95   0.000     .5328072    .5924947
            _cons |   .5534209   .0323615    17.10   0.000     .4899935    .6168483
  ----------------+----------------------------------------------------------------
  r13cesdnmv2     |
      r12cesdnmv2 |   .5448257   .0153369    35.52   0.000      .514766    .5748855
            _cons |   .6105855   .0334158    18.27   0.000     .5450917    .6760793
  ----------------+----------------------------------------------------------------
  r14cesdnmv2     |
      r13cesdnmv2 |   .5806416   .0149033    38.96   0.000     .5514317    .6098516
            _cons |   .5624245   .0333205    16.88   0.000     .4971174    .6277315
------------------+----------------------------------------------------------------
 var(e.r9cesdnmv2)|   3.270004   .0827385                      3.111794    3.436256
var(e.r10cesdnmv2)|   2.248683   .0568968                      2.139887     2.36301
var(e.r11cesdnmv2)|   2.141471   .0541841                      2.037862    2.250347
var(e.r12cesdnmv2)|   2.283458   .0577767                       2.17298    2.399553
var(e.r13cesdnmv2)|   2.411332   .0610122                      2.294667    2.533929
var(e.r14cesdnmv2)|   2.349016   .0594354                      2.235366    2.468444
-----------------------------------------------------------------------------------
LR test of model vs. saturated: chi2(15)  =   1850.63, Prob > chi2 = 0.0000
How can I use margins to predict r9cesdnmv2 by the two values of relwmo2v2c which is a binary variable of 0 and 1? I also want to do this for r10cesdnmv2-r14cesdnmv2 as relwmo2v2c has indirect effect on those variables.

Thanks!

Alice