I have a practical question. I have a big dataset that I have extracted form a language corpus in which I have marked lots of nominal factors. I need to find out how two types of speakers (child vs. adult) are using tow types of structures (prenominal vs. postnominal) in two context subgroups (unmarked vs. marked). Now, none of these data is numerical and I don't know how to procede.
Previously I have extracted the relevant number of occurrences for each combination of factors, made a csv file and inserted it into R studio, but I don't think this gives me enough precision as I then only have the final numbers and I cannot add the random effects such as individual speakers or the occurrence itself.
So what is the recommendation of doing such a 3-way interaction but also keeping in mind random effects?