How to obtain pairwise p-values after linear mixed model

[Al Bothwell]

I run this linear mixed model:

xtset id2 time
mixed sbp ib6.group##i.black_or_not i.stratum age ib2.sex dbp || id2: , cformat(%6.1fc)
margins i.group#i.black_or_not, cformat(%6.1fc)

This is the result:

 Mixed-effects ML regression                         Number of obs    =   5,875
Group variable: id2                                 Number of groups =     890
                                                    Obs per group:
                                                                 min =       1
                                                                 avg =     6.6
                                                                 max =       7
                                                    Wald chi2(15)    = 4687.69
Log likelihood = -21252.326                         Prob > chi2      =  0.0000

---------------------------------------------------------------------------------------
                  sbp | Coefficient  Std. err.      z    P>|z|     [95% conf. interval]
----------------------+----------------------------------------------------------------
                group |
          Acebutolol  |       -4.0        1.1    -3.79   0.000         -6.1        -1.9
          Amlodipine  |       -4.3        1.0    -4.20   0.000         -6.4        -2.3
      Chlorthalidone  |       -5.8        1.1    -5.47   0.000         -7.8        -3.7
           Doxazosin  |       -2.2        1.0    -2.13   0.033         -4.2        -0.2
           Enalapril  |       -3.1        1.0    -2.95   0.003         -5.1        -1.0
                      |
         black_or_not |
               Black  |       -0.5        1.4    -0.36   0.718         -3.2         2.2
                      |
   group#black_or_not |
    Acebutolol#Black  |        2.8        2.3     1.20   0.230         -1.8         7.3
    Amlodipine#Black  |        0.5        2.7     0.20   0.838         -4.7         5.8
Chlorthalidone#Black  |        4.8        2.3     2.11   0.035          0.3         9.2
     Doxazosin#Black  |        0.5        2.4     0.23   0.821         -4.1         5.2
     Enalapril#Black  |        3.8        2.3     1.65   0.098         -0.7         8.4
                      |
            2.stratum |        0.3        0.6     0.42   0.674         -0.9         1.4
                  age |        0.6        0.0    12.21   0.000          0.5         0.6
                      |
                  sex |
                Male  |       -2.9        0.6    -4.67   0.000         -4.1        -1.7
                      |
                  dbp |        1.1        0.0    66.34   0.000          1.1         1.2
                _cons |        9.3        3.0     3.06   0.002          3.4        15.3
---------------------------------------------------------------------------------------

------------------------------------------------------------------------------
  Random-effects parameters  |   Estimate   Std. err.     [95% conf. interval]
-----------------------------+------------------------------------------------
id2: Identity                |
                  var(_cons) |       63.8        3.5          57.3        71.0
-----------------------------+------------------------------------------------
               var(Residual) |       59.3        1.2          57.0        61.7
------------------------------------------------------------------------------
LR test vs. linear model: chibar2(01) = 2430.37       Prob >= chibar2 = 0.0000

. margins i.group#i.black_or_not,  cformat(%6.1fc)

Predictive margins                                       Number of obs = 5,875

Expression: Linear prediction, fixed portion, predict()

-------------------------------------------------------------------------------------------
                          |            Delta-method
                          |     Margin   std. err.      z    P>|z|     [95% conf. interval]
--------------------------+----------------------------------------------------------------
       group#black_or_not |
    Acebutolol#Not Black  |      123.9        0.8   147.39   0.000        122.3       125.6
        Acebutolol#Black  |      126.2        1.6    76.64   0.000        123.0       129.4
    Amlodipine#Not Black  |      123.6        0.8   152.47   0.000        122.0       125.2
        Amlodipine#Black  |      123.6        2.2    57.03   0.000        119.4       127.9
Chlorthalidone#Not Black  |      122.2        0.8   144.64   0.000        120.5       123.8
    Chlorthalidone#Black  |      126.4        1.6    80.70   0.000        123.3       129.5
     Doxazosin#Not Black  |      125.7        0.8   153.12   0.000        124.1       127.3
         Doxazosin#Black  |      125.8        1.8    71.67   0.000        122.3       129.2
     Enalapril#Not Black  |      124.8        0.8   149.91   0.000        123.2       126.5
         Enalapril#Black  |      128.2        1.7    76.12   0.000        124.9       131.5
       Placebo#Not Black  |      127.9        0.6   200.74   0.000        126.7       129.2
           Placebo#Black  |      127.4        1.2   102.83   0.000        125.0       129.9
-------------------------------------------------------------------------------------------

How do I obtain p-values for specific group by race? For example, how do I get the p-value for Acebutolol#Not Black vs. Acebutolol#Black?

Your help will be appreciated. Thank you!

[Clyde Schechter]

You already have them in the -mixed- output. For the placebo group, the black vs non-black contrast is given by the black_or_not output row. For the other groups, you find them in the corresponding interaction row. For example, for Acebutolol#Black vs Acebutolol#Not Black, the results are in the Acebutolol#Black row of the group#black_or_not section of the -mixed- output.

Note that -margins- is irrelevant for this.

[Al Bothwell]

Thank you Clyde. This is very helpful.