工程测量系统分析讲义.pptVIP

The appropriate answer is the shipping scale. The analytical balance is over-kill for the packages being measured and the truck scale does not have enough resolution in only measuring to the nearest 100 pounds. A transactional example of a data collection form with no slot to log in a critical piece of information --Date received. Another example of a gauge with not enough resolution to measure what is requested. Key point: Even commercial gauges (checksheets, audit forms, surveys) need to be checked for proper resolution. Scenario : The issue is - was the shipment correct Customer identifies 25 “bad” shipments in past quarter Customer Service Rep, upon communication with the customer, logs “bad” shipments as follows (see page) Ask the class - What are most of the defects due to ? Does this data collection “gauge” have enough resolution (high enough discrimination Index) to measure the defect in order to reduce the defect ? Answer : NO - Most of the data is classified in the “OTHER” category Now let’s do this exercise to explore the difference between accuracy and precision. Have the participants complete the exercise. I. Accurate and Precise II. Not Accurate, Precise III. Accurate, Not Precise IV. Not Accurate, Not Precise Discuss how they would define accuracy and precision. Accuracy is related to getting the true value. # III is accurate because on average, the observed value is equal to the true value. Precision is related to getting the same or similar values over and over again. In most gauge RR studies, we are looking at precision. A measure of accuracy is bias. Bias effects include : Operator Bias : different operators get detectably different averages for the same thing Machine Bias : Different machines get detectably different averages for the same thing Left gauge: This example shows poor linearity. Notice the number of “positive” responses vs.. the number of negative responses. There are seven that reflect a good rating