# SPSS - MIXED Specifying covariate at level 2 or higher?

#### bramv

##### New Member
Hi there,

Hopefully I'm not posting a stupid question, but here goes:

I'm reading up on doing multilevel analyses in SPSS and following some 'tutorials' and there is one thing that seems weird to me:

As far as I can tell, you do not specify fixed effects for level 1 any different than for level 2 (or 3 etc.). It seems you just have to add the variable name to the FIXED argument, and that's it. Is that correct?

Very simple example: measurements done on students in classes with random intercepts only:

dv = dependent variable
level1cov = covariate measured at level 1 (student level)
level2cov = covariate measured at level 2 (class)

MIXED dv WITH level1cov level2cov
/FIXED = level1cov level2cov | SSTYPE(3)
/METHOD = REML
/PRINT = SOLUTION
/RANDOM INTERCEPT | SUBJECT(class) COVTYPE(VC).

So first question: is this the correct way of specificying the fixed effects for both levels?

And second question: if this is correct, how does the model 'know' on which level the variable is measured? Or does it not need to know that? I guess I don't fully understand... can someone explain a little bit about this? Thanks!

##### New Member
Hi all,

I am facing exactly the same problem. Does anyone have an idea? Or did you, @bramv , figure it out already?

Any assistance would greatly be appreciated, thanks!

#### spunky

##### Doesn't actually exist
SPSS (much like R, STATA and SAS) neither use nor recognize this "Level 1" vs "Level 2" language. That's a direct consequence of the Raudenbush and Bryk software and the whole notation/framework they developed for mixed-effects models. If you want to use SPSS to fit a mixed-effects model, you need to know how to specify a mixed effects equation of the form $$y = X\beta + Zu +\epsilon$$ where the $$X\beta$$ are the fixed effects and $$Zu$$ are the random effects.