Akira Hososhima, CEO of Torin Consulting
Partition analysis is extremely effective when analyzing production data to determine the cause of defects. The partition report stratifies the range in which good and defective products occur, and the corresponding leaf report reveals the number (rate) of defective products (good products) occurring in the variable range, so the conditions for the occurrence of defective products can be clearly understood. The results of this analysis can be understood even without statistical knowledge, so it is easy to explain the analysis results to customers and key people. Another advantage is that interactions are revealed without consciously incorporating interaction terms into the model, as in regression analysis. Of course, partition analysis is not omnipotent. There are cases in which interactions cannot be revealed. It is important to compare profile differences using multiple analysis methods and consider the validity of the model from a technical perspective, not limited to defect cause analysis. Market return prediction can be an important issue that is directly related to management, but the power of JMP's reliability prediction is remarkable. The AICc criterion is shown and the optimal model is revealed, so predictions can be made without wishful thinking. Partial failure models in particular are often useful because they fit the actual situation.
