Stata MIXED Extract Intercept, Slope

[Leon Edelman]

 
 mixed weight week || id: week predict u1 u0, reffects  generate intercept = _b[_cons] + u0  generate slope = _b[week] + u1

If I wish for to estimate and store id-specific intercept and slope I can use the code above, but when I do it with week-squared (week2) I get errors...

 
 mixed weight week week2 || id: week week2  predict u1 u0, reffects  generate intercept = _b[_cons] + u0  generate slope = _b[week] + u1

[Andrew Musau]

Your question is not clear.

when I do it with week-squared (week2) I get errors...

What errors? Use one of the datasets in

help mixed

to illustrate exactly what you are doing and what the issue is.

[Leon Edelman]

Andrew Musau What isn't clear? I specified the model and my issue. The error is that the predicted intercepts and slopes when I include week-squared doesn't make any sense.

[Leon Edelman]

Clyde Schechter Do you have a response to predicting intercepts and slopes when you have a squared-time term?

[Erik Ruzek]

I think the issue might be that you have three random variances but you only ask for two predictions from the predict command.

webuse pig, clear
generate week2 = week*week
mixed weight week week2 || id: week week2
predict u*, reffects
codebook u*, compact
/*
Variable   Obs Unique       Mean        Min       Max  Label
--------------------------------------------------------------------------------------------------
u1         432     48  -1.07e-10  -1.380002  1.192608  BLUP r.e. for id: week
u2         432     48  -7.88e-11  -.1280994  .1339894  BLUP r.e. for id: week2
u3         432     48   4.00e-09  -3.592674  6.074069  BLUP r.e. for id: _cons
*/
generate intercept = _b[_cons] + u3
generate slope = _b[week] + u1 + _b[week2] + u2

[Leon Edelman]

Erik Ruzek but how do i get intercept and slope from u1 u2 u3?

[Andrew Musau]

What isn't clear? I specified the model and my issue. The error is that the predicted intercepts and slopes when I include week-squared doesn't make any sense.

"Doesn't make any sense" isn't meaningful to me. Just like Erik presented a reproducible example, show the output and state what you mean. I would also use factor variable notation here.

webuse pig, clear
mixed weight c.week##c.week || id: c.week##c.week
predict u*, reffects
sum u*

Res.:

. mixed weight c.week##c.week || id: c.week##c.week

Performing EM optimization ...

Performing gradient-based optimization:
Iteration 0:  Log likelihood = -851.98873  
Iteration 1:  Log likelihood = -851.97882  
Iteration 2:  Log likelihood =  -851.9788  

Computing standard errors ...

Mixed-effects ML regression                         Number of obs    =     432
Group variable: id                                  Number of groups =      48
                                                    Obs per group:
                                                                 min =       9
                                                                 avg =     9.0
                                                                 max =       9
                                                    Wald chi2(2)     = 3201.24
Log likelihood =  -851.9788                         Prob > chi2      =  0.0000

--------------------------------------------------------------------------------
        weight | Coefficient  Std. err.      z    P>|z|     [95% conf. interval]
---------------+----------------------------------------------------------------
          week |   6.358818   .1299895    48.92   0.000     6.104043    6.613593
               |
 c.week#c.week |  -.0148922    .012763    -1.17   0.243    -.0399073    .0101228
               |
         _cons |   19.08259   .3943525    48.39   0.000     18.30967    19.85551
--------------------------------------------------------------------------------

------------------------------------------------------------------------------
  Random-effects parameters  |   Estimate   Std. err.     [95% conf. interval]
-----------------------------+------------------------------------------------
id: Independent              |
                   var(week) |   .3755398   .1376915      .1830473    .7704575
              var(week#week) |   .0036763   .0013556      .0017845    .0075734
                  var(_cons) |   5.398871   1.355592      3.300475    8.831398
-----------------------------+------------------------------------------------
               var(Residual) |   1.275934   .1137477      1.071383    1.519538
------------------------------------------------------------------------------
LR test vs. linear model: chi2(3) = 798.37                Prob > chi2 = 0.0000

Note: LR test is conservative and provided only for reference.

.
. predict u*, reffects

. sum u*

    Variable |        Obs        Mean    Std. dev.       Min        Max
-------------+---------------------------------------------------------
          u1 |        432   -1.07e-10    .5025716  -1.380002   1.192608
          u2 |        432   -8.00e-11    .0494083  -.1280994   .1339894
          u3 |        432    4.00e-09    2.142611  -3.592674   6.074069

[Erik Ruzek]

Erik Ruzek but how do i get intercept and slope from u1 u2 u3?

What do you mean by "get intercept and slope from u1 u2 and u3"? The code Andrew and I posted shows you how to get the 48 intercept values for each of the ids. What do you want to do with that information?