Reliability analysis

Reliability refers to the property of a measurement instrument that causes it to give similar results for similar inputs. For example, consider the produce scales at grocery stores. At a given store, each of these scales was probably manufactured at the same factory. You would hope that the factory is reliable - that every scale produced at that factory would register the same weight (within a small margin of error) for the same head of lettuce.

It is important to note that reliability is not just a property of an individual produce scale. When buying groceries, you would expect the particular scale you use to be reliable and to record approximately the same weight when the same item is weighed a second time. However, the reliability of that particular scale is only of immediate importance to the customers using it, while the grocery store is concerned about all of the scales in the store, and the manufacturer is concerned about every scale produced at the factory. The deeper issue here is the reliability of the underlying process of scale manufacture. If that process is reliable, then the manufacturer can be confident that the product is reliable. (Of course, chance error in the manufacturing process will cause a few individual scales to malfunction.)

Worries about reliability are not limited to manufacturers of produce scales but extend to makers of all types of measurement instruments - for example, instructors who write exams for their students, pollsters and marketers who create surveys to gauge public opinion, or trainers who instruct judges for diving meets, beauty contests, or gymnastics competitions. The exams, surveys, and judges' scores are all "scales" that their makers hope are reliable.

However, the produce scale manufacturer likely has a set of standard weights whose exact mass is known; therefore, precise error measurements can be made for the scales. Unfortunately, it is difficult or impossible to establish absolute standards for academic and athletic excellence or the meaning of human responses to a survey. You can only hope to establish scales that are reasonably consistent. The methods available via the Reliability Analysis procedure are useful for situations in which the true state of the measured objects is not known.

Next