The dependent variable, offspring sex, being binomially distributed, was transformed using the logit link function and analyzed assuming a binomial error distribution at the level of the individual.
We therefore maximized the likelihood of the choice proportions (binomial error) and the mean and SD of the RT from correct choices (Gaussian error).
The spot check data for mating rate were modeled using a binomial error distribution as the number of mating opportunities (e.g., the number of females in each treatment × the number of spot checks) taken vs. not taken.
We assumed a binomial error weighted by the total number of eggs laid by each female or each replicate (pair of females).
We analysed differences in knotweed regeneration success using a generalized linear model with binomial error that included the main effects of taxon (3 levels), clone nested within taxon (50 levels), activated carbon (2 levels) and the interactions.
When the GLM revealed a significant interaction between strain and treatment, after significant effects were detected among treatments by the GLM with binomial errors, the method of Benjamini and Hochberg with a false discovery rate [ 53] was applied at the 5% significance level.
We used generalized linear models (GLM) with a binomial error distribution to model the prevalence of Bd in R. pipiens as a function of the following five factors: season, year, stage class, location and tissue-sampling method.
The dependent variable of each model was the response of the bees (binomial error distribution: bees could either land on the inflorescence, 1, or reject it, 0), the explanatory variables were spider position (above or below the inflorescences) and spider movement ("yes" if they moved before the bee landed or "no" if the spider remained still).
GLMs with binomial error distributions (logit) have the advantage of coping with proportion data in a way where sample sizes are taken into account.
A generalized linear model (GLM) with a binomial error distribution tested whether the likelihood of tied versus settled contests depended on sex and RSD between opponents and a likelihood ratio test (LRT) was used to compare nested models.
