Structure and results of the GLMM with binomial error distribution performed in this study.

We performed GLMM with binomial error distribution to compare the probability of a bee visiting flowers with and without slits and with and without nectar within these treatments. As our data set was collect in three consecutive years, we verified a priori that the data did not differ among years by performing the same analysis with the data from each year individually. Here, we show the structure of the models used and the statistical results obtained for each year. These models showed similar results to the final model comprising all the years. Thus, we used for the final analysis our complete dataset, comprising all the sampling years, and added the sampling year as a random variable in the model.