Measurement System Analysis: The Usual Metrics Can Be Noninformative

Abstract

It is important to be able to demonstrate that a measurement process is adequate for the application. Measurement system analysis is a methodology for doing this. The approach is to separate the total variation of a set of measurements into components. The variation among devices being tested is separated from the measurement error. The total measurement error is often separated into several components. Two basic types of metrics utilize the variance component estimates to conclude whether the measurement process is adequate or not. One type of metric compares the measurement error to the variation among tested devices or, equivalently, to the total variation. The other compares the measurement error to the width of the test limits or the product specifications. This paper describes a situation that is becoming increasingly common in testing consumer electronic devices. The situation, called “alignment,” is not actually a “test” in the usual sense, and presents a challenge in using the standard metrics to determine the adequacy of a measurement process. The problem is described and several potential solutions are presented although an ideal solution is yet to be identified.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 15 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.