How can I access model information from multi-level models using the lme4 package?

After running a multilevel model in R using the lme4 package, you may wish to extract information to allow you to look at predicted values, check model assumptions, and understand the results. After running a regression model in R using glm or lm, the returned object is "glm" or "lm". There are numerous R functions (summary, anova, wfit, residuals, coeff, effects) that can be used with these objects. After running a regression model using lmer in the lme4 package, the returned object is "mer". This object is not compatible with the R functions designed for "lm" and "glm" objects and is generally less transparent than these other model objects.

This page will demonstrate several approaches to the "mer" object that allow interesting components to be accessed and examined.

Example 1: Checking Normality of Random Intercepts

Below, we fit a random intercept model. We may then wish to check the normality of the intercepts.

mlm1 <- lmer(mathach ~ female + (1 | id), data = hsb)
summary(mlm1)

Linear mixed model fit by REML 
Formula: mathach ~ female + (1 | id) 
   Data: hsb 
   AIC   BIC logLik deviance REMLdev
 47064 47092 -23528    47053   47056
Random effects:
 Groups   Name        Variance Std.Dev.
 id       (Intercept)  8.1686  2.8581  
 Residual             38.8501  6.2330  
Number of obs: 7185, groups: id, 160

Fixed effects:
            Estimate Std. Error t value
(Intercept)  13.3450     0.2546   52.41
female       -1.3590     0.1714   -7.93

Correlation of Fixed Effects:
       (Intr)
female -0.350

We can view a list of the random intercepts from the model by entering ranef(mlm1), but this is an object of class "ranef.mer". It is not considered a vector of values.

remlm1 <- ranef(mlm1)
class(remlm1)
[1] "ranef.mer"

hist(remlm1)
Error in hist.default(y) : 'x' must be numeric

We can coerce it to be a vector of numeric values that can then be examined.

remlm1 <- as.vector(ranef(mlm1)[[1]])[,1]
length(remlm1)
[1] 160

class(remlm1)
[1] "numeric"

How can I access model information from multi-level models using the lme4 package? | R FAQ

Example 1: Checking Normality of Random Intercepts

Example 2: Generating predicted values based on fixed effects