I have to calculate the repeatability of the agression scores of male birds (measured on a scale from 0 to 4).
For non-multinomial data I would have used a mixed model (lmer) like this:
lmer(Response~1+(1|ID))
And use the residual and interindividual variance to calculate the repeatability.
However, this formula doesn't work for multinomial data.
Any solutions?
Depending on the number of replications, you can run either the two sample t test or the ANOVA test to compare means. If comparison resulted in significant difference, then not repeatable, otherwise repeatable.
Ordinal package (https://cran.r-project.org/web/packages/ordinal/index.html) and clmm function would run that model when your response is an ordinal variable. You should be able to get the residual and interindividual variance also in cumulative link models. In addition MCMCglmm should run that model and have offer information needed to calculate repeatability.
There is also rptR (http://rptr.r-forge.r-project.org/) for repeatability estimation for Gaussian and non-Gaussian data that handles gaussian, binomial and count data. but that probably is not exactly what you need.
- Badges
- Science method
- Similar topics
- Biological Science
- Ecology