Statistical analysis
To understand how social condition influences water conservation, we fit
two linear regression models, one with the absolute mass lost as the
response variable, and one with the proportional mass lost as the
response variable. As predictor variables, we included factors we
expected to influence water conservation, including: 1) the social
condition (singleton vs. paired), 2) the initial body mass, 3) the
proportion of active time, and 4) the interaction between the social
condition and the proportion of active time. The interaction term tested
the hypothesis that social condition influences active time, i.e., if
proximity to another bee stimulates activity. We confirmed that models
met assumptions of normality and homoscedasticity by evaluating QQ-plots
and plots of fitted values versus residuals. We then performed Type-II
ANOVAs on fitted models to evaluate the significance of our predictor
values. We excluded from our analysis any bees that died over the course
of the assay, as well as any bees that were paired with bees that died.
Results are presented as mean ± standard error. All analyses were
performed in R version 4.2.1 using the “car” package [29,30].