I want to decompose the variation of what I characterize as "concordance scores" by applying the mixed command to a complex data structure. I am seeking help regarding the most computationally efficient syntax for the model. I use Stata 15.0

Let rating jekts represent the performance rating of teacher j during observation k in year t by evaluator e in school s, and let rating jjkts represent the performance rating self-reported by teacher j. Define concordance j.kts = rating jekts - rating jjkts

Data structure​​​​​​
  • observations (k) are nested withing years (t), and years (t) are nested within teachers (j ; i.e., teachers receive multiple observations each year and there are multiple years of data)
  • teachers (j) are crossed with evaluators (e ; i.e., teachers are observed by multiple evaluators)
  • teachers (j) and evaluators (e) are nested within schools (s)
I am interested in decomposing variation by school (s), evaluator (e), teacher (j), evaluator-by-teacher (e*j), evaluator-by-year (e*t), teacher-by-year (j*t), and evaluator-by-teacher-by-year (e*j*t).

Please assume that the data meet all modeling assumptions and that estimating these variances is appropriate.