PH—Postharvest Technology: Non-linear Multiple Regression Models to estimate the Drop Damage Index of Fruit

The aim of the study was to determine the best reliable statistic model among linear and non-linear multiple regressions, to calculate an index of the fruit bruising susceptibility estimation that is the drop damage index (DDI). The DDI is determined by a specific numerical procedure where quantity damage estimation is transformed as a probability of damage occurrence, providing a single value that represents the drop height threshold for which the probability of damage is equal to a fixed value (5%). It is based on multiple regression analysis, considering the contribution of different fruit (physical–morphological) and impact (direct drop on rigid support) variables on the average damage. A modified (segmented curve or spline) linear regression and multiple non-linear models were compared in terms of a specific statistical performance score, using a specific fruit and cultivar (apple, Golden clone B). The highest statistically performing model presented the logarithmic algebraic operator in base of ten (log10) applied to both the drop height (H) and firmness (D) variable, without the constant: log10H+log10D. The DDI, calculated for the best model by a numerical procedure, estimated a drop height of 15 mm related to an apple damageability of 5%. The other logarithmic models, obtained by the application of both the operator log10 and lne, presented lower values of the statistical performance score, followed by the square root models and lastly by the linear spline equation. The values of the DDI predicted by the logarithmic models are considered to be more fitting to the impact reality and to cultivar-specific damage sensitivity.