I have a random intercept and random slope mixed linear model with a continuous dependent variable. To calculate ICC, proportion of dependent variable explained by model with predictors at Level-2 and Level-1 I produce "intercept only" Null Model 1 and Null Model 2 (random intercept and random slope, i.e. with time included).

1) Does Intercept-Only Null Model 1 intercept variance .412*** (.089 SD) set the limit on proportion of dependent variable explainable at Level-2 by the full model with predictors and random slopes and intercepts 1.700*** (.378 SD)?

2) Specifically what I am asking: is that .412* intercept variance the only explainable at Level-2 part of 1.700* intercept variance**** in the full mixed-effects model?

Because if I try to compute proportion of dependent variable variance explained at Level-2 by the full model using the formula from Sampson and Bryk (2002, p.74) (intercept variance Null model – intercept variance Model with Predictors) / intercept variance Null model I get = 3.1 i.e. more than 1 or more than 100% (and that never happened in my experience). Is that 3.1 proportion of explained dependent variable variance an anomaly?

3) Should I use Intercepts and Slopes Null Model 2 as the baseline (not Null Model 1) to compare full models with random intercepts and slopes and predictors for ICC, L-1, L-2 calculations with intercept variance 1.713*** (.372)? But that boosts ICC to .79.

I would appreciate any insight!


closed as unclear what you're asking by Michael Chernick, user158565, mdewey, Peter Flom Jan 31 at 10:46

Please clarify your specific problem or add additional details to highlight exactly what you need. As it's currently written, it’s hard to tell exactly what you're asking. See the How to Ask page for help clarifying this question. If this question can be reworded to fit the rules in the help center, please edit the question.