Social behaviour at the beginning of life: the role of quality signals and family size
Description
Data from a experiment in which we manipulated the brood size and the UV/yellow colour of blue tit nestlings to analyze their effect on early sociability. The data is in Excel (.xlsx) format and there is a README document explaining the details of the variables.
Files
Steps to reproduce
Our model system is the blue tit. First, we manipulated the UV reflectance of yellow breast feathers, a sign of quality, by applying an Edding 4500 (code 005) T-shirt marker to 11-day-old nestlings. Second, we manipulated clutch size by exchanging nestlings between nests of similar clutch size and hatch date. Next, we recorded social interactions between nestlings (direct physical contact) with a night-vision video camera. We used this information to create interaction matrices from which we extracted social metrics of interest using UCINET software (Borgatti et al., 2002) and R (R Core Team, 2020). The statistical analyses were carried out in R (v 4.1.1, R Core Team, 2020). The lme4 package (Bates et al., 2015) was used to run generalized linear mixed models (GLMM) with Poisson distribution (log link function) for node degree centrality and linear mixed models (LMM) with normal distribution for the other variables (network degree centrality, node and network average strength, and UV assortment coefficient). In the Poisson model, we checked for overdispersion of residuals following Bolker et al. (2017), which was not the case. Nest ID was included as a random factor in all the models. Nestling ID was included as an additional random factor in the analysis of node degree centrality, for which there were on average 10 values per individual, but not for node average strength (a weighted metric), for which there was one average value per nestling. For node-level analyses, brood size manipulation (enlarged vs. reduced brood size), UV treatment, the interaction between treatments, original brood size, and recording date were included as fixed factors. For the network-level metrics, brood size manipulation, original brood size and recording date were included as fixed factors. Non-significant interactions were dropped from full models. The ANOVA table was obtained using the car package (Fox & Weisberg, 2019), applying Χ2 test for the GLMM model and F tests for the LMM models.
Institutions
Categories
Funding
Ministerio de Economía y Competitividad
PID2019-106032GB-I00