R Statistics Cookbook

How to do it...

Computing the cAIC for mixed models is a nontrivial issue, but fortunately it can be done with the cAIC4 package. In this recipe we will formulate seven versions of our model for deal sizes; we will also add some noise variables into some of the models. If cAIC4 works as expected, it should penalize those models more than the rest:

1. First, we load the necessary libraries and dataset and then we generate four new variables that will have just noise. Ideally models having them, should receive a greater penalization:

library("lme4")  library("glmmLasso")  library(cAIC4)  set.seed(25)  data = read.csv("C:\\R_book\\sample_random_regression.csv")  data$clientid = as.factor(data$clientid)  data$ERR_1 = rnorm(100,0,10)  data$ERR_2 ...