r(1400) after meglm

[Marc Jarczok]

Hi together,

I tried to use meglm models according to this tutorial:
https://stats.idre.ucla.edu/stata/faq/how-can-i-estimate-effect-size-for-mixed/

to calculate effect sizes for a repeated measure analysis. However, with my null model I get this error message:

meglm nil6_pgml || ID:  , constraints(1) 

Fitting fixed-effects model:

Iteration 0: log likelihood = -352.33734  
Iteration 1: log likelihood = -352.33734  

Refining starting values:

Grid node 0: log likelihood = .
Grid node 1: log likelihood = -346.7217
Grid node 2: log likelihood = -346.93043
Grid node 3: log likelihood = -381.39081

Fitting full model:

initial values not feasible
r(1400);

All other previous models were calculated. I have tried using the option:

startgrid()

to set other start values, but the error message remained.

Could one of you explain what the error means, so what the problem is? And you may have a suggestion how to set a value for grid node 0 that is not 0 or . (missing)?
Thanks! ~Marc

Here is the same model but showing the starting values chosen by default:



meglm nil6_pgml || ID: , constraints(1) noestimate

Fitting fixed-effects model:

Iteration 0: log likelihood = -352.33734
Iteration 1: log likelihood = -352.33734

Refining starting values:

Posting starting values:

Mixed-effects GLM Number of obs = 208
Family: Gaussian
Link: identity
Group variable: ID Number of groups = 46

Obs per group:
min = 2
avg = 4.5
max = 10

Integration method: mvaghermite Integration pts. = 7

( 1) [var(_cons[Code])]_cons = -.1122804
---------------------------------------------------------------------------------
nil6_pgml | Coef. Legend
----------------+----------------------------------------------------------------
_cons | 2.312772 _b[nil6_pgml:_cons]
----------------+----------------------------------------------------------------
ID |
var(_cons)| -.1122804 _b[var(_cons[Code]):_cons]
----------------+----------------------------------------------------------------
var(e.nil6_pgml)| 1.733222 _b[var(e.nil6_pgml):_cons]
---------------------------------------------------------------------------------
Note: The above coefficient values are starting values and not the result of
a fully fitted model.

[Carlo Lazzaro]

Marc:
googling with the string -initial values not feasible Stata- gives back many promising entries.