Abstract:
For apples the chain from grower to consumer is divided in different phases: the growing and harvesting period, the storage of the fruits, the sorting and the distribution phase.
The application of non-destructive methods (the weight, the stiffness factor, the background colour and the percentage blush) for monitoring post-storage quality of apples requires adapted statistical methodologies.
The possibility of measuring each fruit several times during shelf life makes the statistical analysis a “repeated measures” one.
This offers important advantages for the testing of treatment effects.
These effects, such as the picking time or maturity, are evaluated in a split plot analysis and tested for their significance in a multivariate analysis of variance.
The impact of maturity (defined in a commercial context as the time of picking) on the quality of apples after storage is evaluated as an example of the statistical methodologies necessary to deal with non-destructive measurements repeated over time on the same fruits.
Apples are picked at random.
This implies that the variability within one pick is seen as a random factor including all types of orchard and tree effects.
This diverges from the actual practice where trees are picked recurrently.
The experiment aims to give insight in the relative importance of four factors: the degree of maturity, the time spent by the product after removal from storage, the interaction between time and maturity and the average random apple variability within one pick.
These fixed and random effects are estimated in the split plot model.
Because of the serial correlation between measurements in time, multivariate analysis of variance models are built to test for the time effects.
Biplot displays are used to describe the evolution of the overall quality of the fruit measured by different non-destructive methods.
|