3
$\begingroup$

I originally learned about random effects models when taking a course on Hierarchical Linear Models, which was taught using Raudenbush and Bryk's HLM book and software, and it sort of indoctrinated me to the subject in a very narrow way. I have a hard time thinking about random effects models in a way other than "Level-1 formula is..." and "Level-2 formula is..." Now I'm trying to migrate away from the HLM software and use instead lme4. But of course even if I specify the model correctly lme4 will give me slightly different estimates than the HLM software, so I'm not 100% sure that I'm doing this correctly. Could someone please tell me whether I am specifying the following model (from chapter 2 of the HLM help manual) correctly?

Level-1 model:

$$MATHACH_{ij} = \beta_{0j} + \beta_{1j}SES_{ij} + r_{ij}$$

Level-2 models: $$\beta_{0j} = \gamma_{00} + \gamma_{01}SECTOR_j + \gamma_{02}MEANSES_j + u_{0j}$$ $$\beta_{1j} = \gamma_{10} + \gamma_{11}SECTOR_j + \gamma_{12}MEANSES_j + u_{1j}$$

This yields the following mixed model:

$$MATHACH_{ij} = \gamma_{00} + \gamma_{01}SECTOR_j + \gamma_{02}MEANSES_j + \gamma_{10}SES_{ij} + \gamma_{11}SECTOR_jSES_{ij} + \gamma_{12}MEANSES_jSES_{ij} + u_{0j} + u_{1j}SES_{ij} + r_{ij}$$

But seeing it in mixed model form doesn't quite help me with making sure I'm doing it right in lme4. Here's what I have so far:

lmer(mathach ~ sector + meanses + ses + sector:ses + meanses:ses + (1+ses|school),data=dat)

We are measuring children (level 1) nested in schools (level 2). ses is a child-level predictor, while sector and meanses are school-level predictors. Subscript i is for children and subscript j is for schools.

$\endgroup$
3
$\begingroup$

Could someone please tell me whether I am specifying the following model (from chapter 2 of the HLM help manual) correctly?

Looks right to me.

$\endgroup$
  • $\begingroup$ I agree. To mee, too. $\endgroup$ – Henrik Sep 17 '14 at 12:23
2
$\begingroup$

If you were using MLWin, you need to add REML=FALSE to the end of your lmer function to get the exact same numbers. Lme4 defaults to REML=T, while MLWin defaults to REML=F.

$\endgroup$

Your Answer

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

Not the answer you're looking for? Browse other questions tagged or ask your own question.