Data considerations for Evaluate Measurement Process (EMP Crossed)

To ensure that your results are valid, consider the following guidelines when you collect data, perform the analysis, and interpret your results.

Operators should measure parts in a random order
To ensure that the data collection order does not influence the results, each operator should measure all parts randomly within a replicate. After all operators measure all parts one time, you repeat the process for all replicates.
Operators should measure at least 10 parts for an adequate study
The process variation can be estimated from a large sample of historical data, or from the parts in the study. If you have a historical estimate of the process variation, the usual requirement of 10 parts is acceptable.
If you do not have a historical estimate, you should consider using more than 10 parts. Although you need a large number of parts to obtain a very precise estimate of process variation, using 15 to 35 parts should result in a much better estimate than if you used 10 parts.
Select parts that represent the actual or expected range of process variation
Select parts from the entire range of your process to increase the likelihood of having a good estimate of the part-to-part variation. For example, do not measure consecutive parts, parts from a single shift or a single production line, or parts from the reject pile.
Operator and Part factors must be crossed
Two factors are crossed when each level of one factor occurs in combination with each level of the other factor.
For example, if the Operator and Part factors are crossed, every operator must evaluate every part.
For more information on crossed factors, go to Types of factors in gage R&R studies and Wheeler's EMP studies.
Operator and Part factors must be random
A factor is random when the factor has many possible levels, but only a random sample of the levels is included in the data.
For example, Part is a random factor because the parts selected for the study are meant to represent all possible parts from the production process.
For example, Operator is a random factor when you have many employees who take measurements, but you randomly select a few operators for the study.
For more information on random factors, go to Types of factors in gage R&R studies and Wheeler's EMP studies.
Operators must measure each part at least twice
Measurement variation is broken down into two components: reproducibility and repeatability. Reproducibility is the variation that occurs when different people measure the same part. Repeatability is the variation that occurs when the same person measures the same part repeatedly. If you use at least 10 parts and at least 3 operators, having each operator measure each part at least 2 times, in random order, allows you to obtain an adequate estimate of repeatability.
You should have at least 3 operators for an adequate study
For the best results, include 3 to 5 operators in your study. You should not have fewer than 3 operators in the study, unless the number of operators who use the measurement system is actually less than 3. If you suspect that there are large differences between operators, you should consider using more than 3 to 5 operators. If you identify differences between operators, such as an operator whose measurements are lower than other operators, you can often improve consistency with training.
When selecting operators for the study, ensure that they are representative of all the operators who use the measurement system. If you perform the study with only the best (or worst) operators, the results will be biased and will not provide an accurate estimate of operator differences. The best way to ensure accuracy is to randomly select the operators for the study.