Akira Hosojima, President, Tobayashi Consulting
When analyzing production performance data to determine the cause of defects, the partition analysis shows outstanding sharpness.The partition report stratifies the range where non-defective products and defective products occur, and the corresponding leaf report reveals the number (rate) of defective products (non-defective products) in the variable range, so the condition of defective products can be clearly understood. can do.Since the analysis results can be understood without statistical knowledge, it is easy to explain the analysis results to customers and key persons.It is also an advantage that the interaction becomes clear without intentionally incorporating the interaction term into the model as in regression analysis.Of course, partition analysis is not all-purpose.Sometimes the interaction cannot be revealed.Although it is not limited to failure cause analysis, it is important to compare the differences in profiles using multiple analysis methods and to examine the validity of the model from a technical point of view.Market return forecasting can be an important issue directly linked to management, but JMP's reliability forecasting power is striking.Since the AICc criterion is shown and the optimal model is clarified, it is possible to make a prediction without the desired bias.In particular, the partial failure model is often useful because it fits the actual situation.
