I have a data set of animal responses with repeated measures. The responses are on a 9-point scale. Here, I present an example data set that is very similar to mine:



# number of subjects
n <- 300

# number of measurements per subject
k <- 4

# Assign a unique 4-digit ID number to each subject
rep(sample(1111:9999, n, replace = FALSE), each = k) %>%
  as.factor() -> id

# Generate random responses from subjects
c(replicate(n, c(sort(sample(1:9, 4, replace = TRUE))))) %>%
  as.numeric() -> score.numeric
  score.numeric %>%
  as.ordered() -> score.factor

# Assign sex to subjects
rep(gl(2, n/2, labels = c("male", "female")), each = k) %>%
  as.factor() -> sex

# Assign a generation to subjects
rep(gl(2, n/2, labels = c("p", "f1")), each = k) %>%
  as.factor() -> gen

# Mix the assignments within "gen"
gen <- sample(gen)

var.list <- list(id = id,
                 score.numeric = score.numeric,
                 score.factor = score.factor,
                 sex = sex,
                 gen = gen)

DF <- do.call(cbind.data.frame, var.list)

I employed the R package "ordinal":

model.clmm <- clmm(score.factor ~ sex + gen + (1 | id),
                   link = "cauchit",
                   Hess = TRUE,
                   data = DF)

I have two main questions:

  1. Is it possible to obtain a repeatability value (ICC) with a 95% confidence interval using this package?

  2. How do I analyze the fit of this model? I am new to ordinal models in general. My approach in the past with mixed models was to test the residuals for normality. It seems I cannot do the same with a clmm() model, or at least, doing so is not immediately obvious to me.


1 Answer 1


If you mean intraclass correlation coefficient when you mention ICC, as far as I know you can only calculate it by yourself as of today. According to Liu,2005, p.414, ICC= τ/(τ+3.29), where τ is between-group variance.

As of the assessing model fit part, you can provide -2LL, LR, delta(df), AIC, BIC, SABIC, log-likelihood, SABIC values, and also MuMIn Multi-Model Inference R package provide AICc value and model comparision functionality for ordinal clmm objects.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.