Comparative performance of three test methods for determining chipping quality in tubers of potato (Solanum tuberosum L.)

The capacity of a colour chart and a reflectance photometer (Agtron) to acc urately determine chipping quality of potato tubers was assessed using data sets taken over a 4-yr period for 17-32 cultivars. Both tests gave a high diagnostic accuracy for chipping quality regardless of sampling time from s torage or the occurrence of high temperature reconditioning when evaluated by receiver operating characteristic (ROC) curve analysis. Receiver operati ng characteristic curve analysis showed that both tuber glucose content and chip colour provided good diagnostic performance in correctly separating p rocessing from non-processing tubers over a range of growing and storage co nditions. Identification of chipping from non-chipping tuber samples From a 13 degrees C storage across a range of cultivars and growing conditions oc curred with a minimum chipping colour threshold range of 41-47 or a maximum glucose concentration range of 4.3-5.4 mmol L-1 of tuber cell sap. The pra ctical value of a test can depend on such factors as prevalence of chippers in a tuber population as well as the cost of misclassifications, i.e., cos ts associated with false positive or false negative test results and expres sed in relative terms as the unit cost ratio. An examination of Prevalence- Value-Accuracy (PVA) plots for one of the data sets indicated that total mi sclassification costs could increase rapidly, depending on the prevalence o f chipping tubers and the relative amounts of false negative and false posi tive costs. Maximum costs were consistently associated with a prevalence of 50% chippers and a unit cost ratio of 0.5. In a tuber sample containing a high prevalence of chippers (50-70%) and a low unit cost ratio (<0.2), an a cceptable colour threshold determined by PVA-Threshold (PVAT) plots would b e approximately 40 to 50 from the Agriculture and Agri-Food Canada colour c hart. However, if the colour chart was used for screening tuber samples wit h a low prevalence (20-40%) of chippers and a unit cost ratio >0.20, a thre shold between 60 and 65 would be optimum. The latter range would be conserv ative and agrees with, and supports, current industry standards, which resi de at 60 or better. Since a good diagnostic test should be repeatable and s ubject to minimal inter-observer variation, the more objective glucose or r eflectance photometric tests may be preferable and provide acceptable diagn ostic accuracy for processing quality. However, the present study indicates that all three test methods are acceptable for accurately separating chipp ing from non-chipping tubers regardless of sampling or storage protocols.