I have a simple question, understanding the basic usage of the

`lme4`

package. I am following the tutorial by Bodo Winter (http://www.bodowinter.com/tutorial/bw_LME_tutorial.pdf).In this tutorial, Bodo calculates a random effects model using the two commands:

`library(lme4) politeness=read.csv("http://www.bodowinter.com/tutorial/politeness_data.csv") politeness.model = lmer(frequency ~ attitude + (1|subject) + (1|scenario), data=politeness) summary(politeness.model)`

However, his printout of the output includes the AIC and BIC values (page 8), which are not included in the current version of lme4 (1.1.7). Do you have any idea why this is the case? Although, one can compute the two values using the maximum likelyhood algorithm (by using the REML=False option), I am confused why they are no longer included in the default output.

Thanks in advance

**Answer**

As far as I can tell this was implemented in Aug 2013 ; the logic would presumably be that models fitted with REML do not have a likelihood *per se*, and that one of the most common user errors is to compare REML criteria (“restricted likelihoods”) across models with different fixed-effect components, which is meaningless. Comparing AIC/BIC would inherit the same problems.

Although lme4 follows a fairly standard R convention of reporting the AIC, BIC, etc. in summary, I actually think this is mostly useless anyway, since the AIC/BIC for a single model basically doesn’t contain any information. You can use it to compare across models, but that’s easier to do with `anova(model1,model2)`

or `AIC(model1,model2)`

(or `bbmle::AICtab(model1,model2)`

, which gives a more useful summary).

**Attribution***Source : Link , Question Author : Andy , Answer Author : Ben Bolker*