In order to evaluate the performance of an existing system, first the performance criteria of the system must be decided upon. Once the criterion is defined, it is necessary to collect data from the system. However, collecting data with varied tools, methods and people would lead to inconsistent results which could result in incorrect conclusions. Even with a standardized method of measurement a measurement error always exists. Measurement System Analysis evaluates the percentage of this error and decides if the measurement system is acceptable or not.
Often used at the measure phase of Six Sigma methodology, Measurement System Analysis (MSA) is a statistical and scientific tool to ensure the measurement done to collect data is consistent, reliable, unbiased and correct. It emphasizes on standardization of data collection method and assessment of the collected data. By doing so the error on the collected data is minimized.
Depending on the type of data, the statistical analysis will be different. For a continuous measurement, there are a variety of statistical properties that can be determined: stability, bias, precision (which can be broken down into repeatability and reproducibility), linearity, and discrimination. For a discrete measurement, estimates of the error rates can be determined for within appraiser, each appraiser versus standard, between appraisers, and all appraisers versus standard.
For discrete measurements, imagine a case where appraisers are required to identify if an inspected object (product) should be classified as OK or not OK according to specified quality standards. In this case a blind study can be done where a number of OK and not OK products are given to two or three appraisers. The appraisers each then determine if they think the product is OK or not OK. They are asked to look at the same unit more than once, without knowing that they had evaluated the unit previously. This is called the “within appraiser” error rate. It can then be determined how well all the appraisers are able to get the same result on the same product, the “between appraiser” error rate. In addition, it can be determined how well the appraisers agree with the expert, known as the “appraiser versus standard” error rate.
For continuous data measurements as it was emphasized before the data is evaluated by the following criteria:
Stability: Corresponds to the ability of the measurement system to produce same results when same sample is measured.
Bias: Is the difference between the actual mean of a sample and its’ measured mean.
Linearity: Shows to what extend is the measurement error has linearity with the measured value. For example, if the measurement value of a 100cm long object has 1cm error, but the error is 5cm on 150cm object using the same measurement system, it could be concluded that the measurement system in not linear.
For determining the variation of the measurement system there are two criteria to that needs to be assessed:
Repeatability: Shows to what extend does the appraiser gets the same results by assessing same sample multiple times using the same measurement system.
Reproducibility: Shows to what extend different appraiser gets the same results on assessing the same sample multiple times using the same measurement system.
After evaluating the measurement system according to above mentioned criteria, if error is below 10 percent the measurement system is considered reliable. If the error is between 10 – 30 percent, the reliability of the system is questionable and improvement of the measurement system is advised where it is applicable. Finally, if the error is above 30 percent, the measurement system is decided as unacceptable and no analysis or conclusion shall be made based on the data collected with such measurement system.
At 6Sigma.us we are committed to helping people find solutions! We provide hands-on implementations of Lean and Six Sigma at our locations, at your workplace or online. Visit our schedule of classes and find a solution that meets your needs, or contact us and we will surely help you find the right fit.