When a higher level can't be modeled due to one row
If measure varies within one study level, but is present and constant in the others, then you can still do: ?(1 | study / outcome) + (1 | measure) You'll just get one BLUP/conditional mode for each study where measure is constant. That's fine and expected -- the measure is constant for the entire study, so its associated random effect shouldn't change. Of course, the measure and study effects (BLUPs) are conflated for every study but 2 (so you shouldn't overinterpret the BLUPs), but you may still be able to separate their overall variance contribution (the bits reported in the "random effects" section of lme4's output). The big fine print here is that you need to have enough - observations in each study - levels of study (with only 3 levels, study probably shouldn't be treated as a blocking variable....) - variabiity between studies - levels of measure in study 2 - variability between measures
On 11/9/21 12:36, Farzad Keyhan wrote:
Dear Colleagues, I initially posted my query on the meta-analysis SIG. But I realized my question is related to multilevel modeling. Below is my data structure. If in row # 3, "measure" was 2 (instead of 1), then, I could model "measure" as a level above "study": (1 | measure / study / outcome) But right now, because in study 2 (rows # 3 and 4) "measure" can vary, "measure" can't be considered a level above "study". On the other hand, because "measure" varies only in one "study" level (i.e., study 2), I can't model "measure" as a crossed effect with "study" either: (1 | study / outcome) + (1 | measure) So, to model "measure" as a random factor what can we do? Best, Fred measure study outcome 1 1 1 1 1 2 # 1 2 1<--row #3 2 2 1 1 3 3 1 3 2 2 4 3 9 5 1 9 6 2
_______________________________________________ R-sig-mixed-models at r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-sig-mixed-models