We are about to measure the quality of a conversation in several ways and have some possible predictors. We would like to have each of 10 people have a conversation with each of the other nine people giving us 45 conversations. I plan to carry out a linear mixed model using lmer(). My problem is that I have two random effects: person 1 and person 2 in each conversation. I could specify two random factors in my model but the problem is that over the course of the experiment the two random variables are the same people and their coefficients will be very highly correlated. Any suggestions?
Asked
Active
Viewed 69 times
2
interaction(person1,person2)) as an additional random effect grouping variable? Or are yourperson1andperson2completely correlated, in which case you should probably specify only the pairs? – Ben Bolker Jan 10 '17 at 17:34