2
$\begingroup$

my question relates to this post: Extracting slopes for cases from a mixed effects model (lme4)

with Sven Hohenstein's great explanation. My question is: how do I extract/interpret individual coefs if I have an additional 2nd level predictor? My random slope model with cross-level-interaction:

Scaled residuals: 
    Min      1Q  Median      3Q     Max 
-5.3837 -0.4603 -0.0878  0.3895  5.1209 

Random effects:
 Groups   Name        Variance Std.Dev. Corr 
 CODE     (Intercept) 0.28950  0.5381        
          MZP.log     0.04025  0.2006   -0.38
 Residual             0.09409  0.3067        
Number of obs: 13866, groups:  CODE, 3917

Fixed effects:
                       Estimate Std. Error         df t value Pr(>|t|)    
(Intercept)           9.925e-01  9.822e-03  3.936e+03 101.055  < 2e-16 ***
MZP.log              -1.771e-01  5.634e-03  4.615e+03 -31.432  < 2e-16 ***
ABS002P1.cgm          5.658e-03  5.767e-04  3.884e+03   9.811  < 2e-16 ***
MZP.log:ABS002P1.cgm  1.192e-03  3.235e-04  3.709e+03   3.686 0.000231 ***


Correlation of Fixed Effects:
            (Intr) MZP.lg ABS002
MZP.log     -0.510              
ABS002P1.cg -0.009  0.041       
MZP.:ABS002  0.045 -0.223 -0.502

I get individual coefs using coef(RS.L2)$CODE which gives:

                          (Intercept)    MZP.log   ABS002P1.cgm  MZP.log:ABS002P1.cgm
==AM0MDMx0CM4ETM0ITZmV2a   1.4243088 -0.1719496549   0.00565832          0.001192448
==AM0UDMx0SO5cDM4EDajVma   1.9458975 -0.0529674147   0.00565832          0.001192448
==AM1cDMx0SN1ATM1ITdnlGZ   0.8161014 -0.1141229315   0.00565832          0.001192448
==AM1ITM0AzV               0.4187989 -0.0886104357   0.00565832          0.001192448
==AM2ADMx0iN5kDMxEjcmVGa   1.1081334 -0.2264506549   0.00565832          0.001192448
==AM2gDMzAjR               0.8997179 -0.2053793890   0.00565832          0.001192448
==AM2IDM0AjQ               0.6851151 -0.2432147380   0.00565832          0.001192448
==AM2kDMwMDajxWZ           0.9268792 -0.0418086190   0.00565832          0.001192448
==AM2UDM3ATYtVGb           0.8382071 -0.1377587548   0.00565832          0.001192448
...

what I don't understand: are the first two columns (Intercept and MZP.Log) a combination of fixed and random effects? I assume yes, but to what part in the final equation

Yti = Beta00 + Beta10 * MZP.Log-ti + Beta01 * ABS002P1.cgm-i + Beta11 * MZP.Log-ti * ABS002P1.cgm-i + r1-i * MZP.Log-ti + r0-i + e-ti

do they refer? Would (Intercept) = Beta00 + r0-i and MZP.Log = Beta10 + r1-i? My aim is to illustrate an individual equation for a specific person (CODE).

I am deeply grateful for any help!

$\endgroup$
5
$\begingroup$

The thing you are missing here is the output of ranef(RS.L2).

This should give you 2 columns of numbers that are the individual random effects for each level of CODE for the intercept and MZP.log

For example, for the first level of CODE (==AM0MDMx0CM4ETM0ITZmV2a) the random effect for the intercept should be be 1.4243088 - 9.925e-01 = 0.4318088

That is the individual intercept for a particular level of CODE is equal to the global intercept plus the random effect for that level of CODE. Likewise for MZP.log

$\endgroup$
1
  • $\begingroup$ Thank you! That's what I was thinking but wasn't sure. $\endgroup$ – 1k2a3l4m Aug 21 '20 at 10:49

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.