*2.4. Data Analysis*

To assess species diversity in each UGA, we calculated diversity indices of Shannon– Wiener (H ) and dominance of Berger–Parker [29]. In addition, we calculated the similarity between UGAs concerning the species composition fauna through the Bray-Curtis index.

The correlation between plant habit and species richness for each guild (related to trophic specialization, nesting behavior and nesting site) was evaluated using a Principal Component Analysis (PCA). These analyses underwent the Past 3.2 program, assuming a 95% significance level [30].

We constructed multivariate generalized linear models (GLMs) with negative binomial distribution to identify the effects of environmental variables (predictors: paved area, DBH of trees, plant richness, flower coverage, paved area around and number of buildings with more than three floors) on richness and abundance of bees in each guild. We tested the collinearity between the predictor variables using the variance inflation factor (VIF) of the car package of the R program. The best model was selected using the lowest value of the Akaike information criterion (AIC). We used the R version 4.0.5 program for these analyses, assuming a 95% significance level [31].
