※ Correlation coefficient and standard error of the calibration model

The predicting performance of a nondestructive measuring instrument can be evaluated by correlation coefficient((R2) and standard error(SEC) of the calibration model.

The correlation coefficient indicates the level of co-relationship between the sugar contents predicted by calibration model and actual sugar contents of extracted juice measured by refractometer. Bigger the correlation coefficient of the calibration model, better the definiteness between high sugar contents and low sugar contents. Thus we can tell the calibration model with bigger correlation coefficient is better optimized to object fruit. The standard error is the statistic distribution of error range between sugar contents predicted by calibration model and actual sugar contents measured from the juice extracted. A standard deviation of 0.5 Brix means that among the total number of measurement, about 68% can predict the sugar contents within the error range of ±0.5 Brix. Thus, in some cases, a measurement error may have larger value than 0.5 Brix. To create a calibration model with good correlation coefficient and useful standard error, it is recommendable to make models collected from various range of sugar contents and considerable number of fruit samples. The correlation coefficient and standard error should be made to be suitable for desired purposes of instrument use ; make decision of harvesting time or selection of favorable fruit at post-harvest And the calibration model should be created in consideration of the varieties of fruit and area/time of cultivation.