*2.4. Statistical Analyses*

General linear mixed-e ffect models for the analysis of variance (ANOVA) were used using R statistical software and the *lmerTest* package to check for the e ffects of treatments and years, after verifying the normality and homoscedasticity of errors. In the case of fruit number, data were modelled in a generalized linear mixed-e ffect model (*lmerTest* package) using Poisson distribution. In all models, treatments (systems) and years were used as fixed factors, and blocks and years as random ones. Data were presented separated by year because of the significant year e ffect and interaction between year and treatment in most of the cases. Pairwise comparisons for all variables were computed by estimating the 95% confidence interval of the di fference between the least squares means (Equation (2)):

$$\text{CI(difference)} = (\mathbf{x}\_1 - \mathbf{x}\_2) \pm 1.96\sqrt{\left(SE\_{x1}\right)^2 + \left(SE\_{x2}\right)^2} \tag{2}$$

where (*x*1) is the mean of the first value, (*x*2) is the mean of the second value, (*SEx*1) is the standard error of (*x*1), and (*SEx*2) is the standard error of (*x*2).

If the resulting 95% confidence interval (CI) of the di fference between values did not cross the zero value, the null hypothesis that the compared values are similar was rejected.

All data in the manuscript were reported in the original scale as least square means with their corresponding standard errors. Results of all analysis of variance/deviance in terms of *p*-values are presented in Table S1.
