Testing whether sex and size contribute to among-individual variation
To assess the extent of the contribution of sex and/or size to the among-individual variation in behaviour, we refitted all models described above but with additional fixed effects of sex (a 3 level factor: male, female, unknown), size (mass as a covariate) and their interaction sex:size. Using univariate models and conditional F-tests, we tested the significance of these effects on each trait. We compared estimates of VI and R from these expanded models to those obtained above. We then refitted our multivariate model with the additional fixed effects and estimatedIDsex:mass , the among individual (co)variance matrix conditional on sex and mass. We scaled covariances to correlations, and subjected IDsex:mass to eigen decomposition as described above, allowing us to compare its structure to ID . We also checked for any clustering of personality visually by sex and/or size in a multivariate personality space, by predicting and plotting individual scores onidmax for each shrimp.