Conducting an MSA Test on scales

[WWscience]

I was required to evaluate whether the current calibrations applied to some scales are allowing for accurate weight readings. I've read that since the scales control an aspect of production (i.e. the quantity of raw materials in a given product), I should conduct an MSA study on the scales. However, this is a very intermediate step in the overall process, and for this reason I had conducted an independent two sample t-test instead, comparing the scale readouts between a scale of interest and a separate/reliable scale. Is the t-test sufficient, or should an MSA test be conducted as well?

 

[Miner]

The two answer very different questions:

• The t-test answers the question: Do the mean weights of the two scales differ from each other?
• The MSA, consisting of:
  • Bias - Answers the question: Does the mean weight differ from the standard.
  • Linearity - Answers the question: Does the bias differ by the magnitude of the measurement?
  • Stability - Answers the question: Does the bias change over time?
  • R&R - Answers the question: What is the precision (variation) when one person makes the same measurement multiple times? AND What is the precision when multiple people make the same measurements?

Notice that these are all different questions.

 

[WWscience]

The two answer very different questions:

• The t-test answers the question: Do the mean weights of the two scales differ from each other?
• The MSA, consisting of:
  • Bias - Answers the question: Does the mean weight differ from the standard.
  • Linearity - Answers the question: Does the bias differ by the magnitude of the measurement?
  • Stability - Answers the question: Does the bias change over time?
  • R&R - Answers the question: What is the precision (variation) when one person makes the same measurement multiple times? AND What is the precision when multiple people make the same measurements?

Notice that these are all different questions.

Thank you for your response. I was thinking that the t-test could be sufficient to answer the question of whether the scale of interest was preforming reliably based on how close the readouts were to the readouts of a scale that's used as a baseline reference. I understand that which direction is chosen is based on the question to be answered. However, if the question is only to determine whether the scale of interest is properly calibrated, would you believe that the t-test is the proper variable to calculate? The scales are calibrated every so often, and the intention behind this study is to ensure that the calibrations are providing for accurate readings every time this instrument is used.

Thank you

 

[Bev D]

One thing to consider is if the weighing system is repeatable. (At this point I am assuming you are weighing a single raw material and not weighing a box full of parts to assure a count in the box) There is more than just the scale at work here. Some raw material can stick to certain ‘scoop’ material. This changes how much material is delivered to the scale. Some material is very ‘slippery’ and can fly away. While this isn’t a ‘traditional’ repeatability gauge R&R it is one of the considerations of a measurement system and goes to the function and purpose of said system. This is why I teach and practice MSA and not just “Gauge R&R”.