I am trying to estimate a three-level model in R using lmer, and I'm a bit confused on a few things, both statistical theory and estimating the model in R.

The data is repeated measures (pre-test vs. post-test, so t=0 or t=1), nested in individuals, nested in programs. I am interested in two things. First, the 40-ish programs fall into 3 "groups". I want to examine the intercept and slope of each group. (The slopes of the groups should show which groups are most/least effective.) Second, I want to know how individual-level covariates affect both intercepts and slopes.

First, this is my interpretation of the statistical equation. Does this look right? Let t $\in [0,1]$ be the repeated measures, $i$ represent an individual, and $j$ represent a program.

Level 1:

$Y_{tij} = \beta_{0ij} + \beta_{1ij} * time_{tij} + error$

Level 2:

$\beta_{0ij} = \pi_{0j} + \Theta_0 * INDIVIDUAL COVARIATES_{ij} + error$

$\beta_{1ij} = \pi_{1j} + \Theta_1 * INDIVIDUAL COVARIATES_{ij} + error$

Level 3:

$\pi_{0j} = \rho_{01} * GROUP1 + \rho_{02} * GROUP2 + \rho_{03} * GROUP3 + error$ $\pi_{1j} = \rho_{11} * GROUP1 + \rho_{12} * GROUP2 + \rho_{13} * GROUP3 + error$

I understand everything except the part dealing with the individual covariates. I do not care about the interaction of program with individual covariates. Is this the right spot to put individual covariates, or am I off base here?

Second, I want to implement this using the lmer package in R. I'm just utterly at a loss on how to do this.

This source (http://rpsychologist.com/r-guide-longitudinal-lme-lmer) discusses a conditional three-level growth model, which seems close, but I do not understand the $(time*tx)$ [I guess $(time*group1)$,$(time*group2)$ etc.] notation at all, nor do I understand how to include individual covariates.

Any help would be deeply appreciated!


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.