I am running a saturated model (-mixed-) for preliminarily assess the effect of intervention for two time points (baseline and time-1).
Outcome variable: meanscore_body (continuous, ranging from 1-5)
Group indicator: studygroup (0: control; 1: intervention)
Time indicator: time (1: baseline; 2: time-1)
I am running the codes:
Code:
global restrict = "t1dropcase == 0 & t2dropcase == 0 & _merge2 == 3" mixed meanscore_body i.time##i.studygroup if $restrict || surid: , residual(uns, t(time)) var ml
Code:
Iteration 270: log likelihood = -5412.7565 (backed up) Iteration 271: log likelihood = -5412.7565 (backed up) Iteration 272: log likelihood = -5412.7565 (backed up) Iteration 273: log likelihood = -5412.7565 (backed up) Iteration 274: log likelihood = -5412.7565 (backed up) Iteration 275: log likelihood = -5412.7565 (backed up) Iteration 276: log likelihood = -5412.7565 (backed up) Iteration 277: log likelihood = -5412.7565 (backed up) Iteration 278: log likelihood = -5412.7565 (backed up) Iteration 279: log likelihood = -5412.7565 (backed up) Iteration 280: log likelihood = -5412.7565 (backed up)
1) What the "backed up" mean in my case?
2) Does long-time of iteration process suggest model mis-specification?
3) How can I modify to get it converged?
Thanks so much in advance!!
Best,
Mengmeng
0 Response to ML estimation goes forever in longitudinal analysis
Post a Comment