mixed and time varying IV

[carole fantini]

Dear all
I ran an experiment where physiological datas for participants (n=147) were taken during a stress task, as well as their level of anxiety at each occasion and a time invariant IV (negative affectivity).
DV : HRV (not transform and not normal)
IV : occasion with 3 levels (categorical); anxiety at each occasion (anx89. continuous and centered at the mean) and affect suppression(AS45.continuous and centered at the mean).
I ran the following model :

mixed HRV i.occasion##c.anx89##c.AS45 || id:,mle cov(un) vce(robust)

and obtain the following results :

. mixed hrv i.occasion##c.AS45##c.anx89 ||id:,mle cov(un) vce(robust)
note: single-variable random-effects specification in id equation; covariance structure set to
      identity.

Performing EM optimization ...

Performing gradient-based optimization:
Iteration 0:  Log pseudolikelihood = -1661.8501  
Iteration 1:  Log pseudolikelihood = -1661.8501  

Computing standard errors ...

Mixed-effects regression                             Number of obs    =    399
Group variable: id                                   Number of groups =    147
                                                     Obs per group:
                                                                  min =      1
                                                                  avg =    2.7
                                                                  max =      3
                                                     Wald chi2(11)    = 206.67
Log pseudolikelihood = -1661.8501                    Prob > chi2      = 0.0000

                                              (Std. err. adjusted for 147 clusters in id)
-----------------------------------------------------------------------------------------
                        |               Robust
                    hrv | Coefficient  std. err.      z    P>|z|     [95% conf. interval]
------------------------+----------------------------------------------------------------
               occasion |
                     2  |  -8.300005   1.597175    -5.20   0.000    -11.43041     -5.1696
                     3  |   8.443544   1.153635     7.32   0.000      6.18246    10.70463
                        |
                   AS45 |   .1817955   .1598155     1.14   0.255    -.1314371    .4950281
                        |
        occasion#c.AS45 |
                     2  |  -.3145298   .1646679    -1.91   0.056    -.6372729    .0082134
                     3  |   .0727873    .100163     0.73   0.467    -.1235286    .2691032
                        |
                  anx89 |  -1.451473   1.221485    -1.19   0.235     -3.84554    .9425945
                        |
       occasion#c.anx89 |
                     2  |   2.057651   1.411346     1.46   0.145    -.7085371    4.823839
                     3  |   .0981878   1.531299     0.06   0.949    -2.903103    3.099478
                        |
         c.AS45#c.anx89 |  -.1681481   .1042543    -1.61   0.107    -.3724828    .0361865
                        |
occasion#c.AS45#c.anx89 |
                     2  |    .303042   .1245039     2.43   0.015     .0590189    .5470652
                     3  |   .2802108   .1363958     2.05   0.040       .01288    .5475417
                        |
                  _cons |   59.60581   1.838499    32.42   0.000     56.00242     63.2092
-----------------------------------------------------------------------------------------

------------------------------------------------------------------------------
                             |               Robust          
  Random-effects parameters  |   Estimate   std. err.     [95% conf. interval]
-----------------------------+------------------------------------------------
id: Identity                 |
                  var(_cons) |   368.2115   42.61907      293.4769    461.9773
-----------------------------+------------------------------------------------
               var(Residual) |   102.4341   12.70524      80.32801    130.6236

I'am interested by the affect suppression effect related to the change of anxiety at each occasion. So my focus is on the

occasion#c.AS45#c.anx89_2

interaction.
After reading some similar question on the statist, I understand that when anxiety is constant, one point increase on affect suppression lead to .30 increase in HRV.
When using margins to compare slopes and mean HRV when AS45 and anx89 are ±1SD above or below the mean, I understand that, in this case, anxiety is average across occasions which is not what I want. Am I correct in this interpretation? if so, how could I test the anxiety increase or decrease between T1 and T2 on the influence of affect suppression on HRV??
thanks a lot for your help.