Locked lesson.
About this lesson
Exercise files
Download this lesson’s related exercise files.
Gage R&R Execution and Analysis.docx61.1 KB Gage R&R Execution and Analysis - Solution.docx
61.2 KB
Quick reference
Gage R&R Execution and Analysis
There are best practices for executing the Gage R&R Study plan and best practices for how to approach the analysis. In addition, there are target performance thresholds for the analysis results.
When to use
The Gage R&R execution best practices should be applied once the Data Collection phase begins. The analysis best practices start at the time of data collection to ensure that the data is measured and recorded correctly. They also apply once the calculation results are completed.
Instructions
Execution Best Practices
The obvious best practice is to follow the Gage R&R Study plan. With regards to data collection, the first appraiser should evaluate all items one time in a random order. The second appraiser evaluates all the items once in a random order and this pattern continues for all appraisers. Once all appraisers have completed their first measurement on all items, the first trial is complete. This process is then repeated for the second trial. The random order for each appraiser is changed for each trial.
The data recording is the most common area of mistakes. Use the form or format of the study plan to ensure that the correct value is recorded for the correct appraiser, item and trial.
Analysis Best Practices
The analysis best practices can actually begin as soon as data collection starts. That is because one of the best practices is to ensure the appraisers are following the plan and data is recorded correctly. This is done as an audit or assessment of the process. The actual calculation for determining repeatability or reproducibility cannot be started until all data has been collected. The calculation approach is different for variable and attribute data and the results are expressed in different terms. But what is the same, is that if the results are unacceptable, the process needs to be improved or a different measurement system needs to be used for that application.
The target performance levels for both types of studies are shown below.
Hints & tips
- Before starting the study, do a “pilot run” of the measurement process and the data recording.
- If the appraisers are not familiar with the Gage R&R process, spend some time with them before they start to inspect to answer their questions and build their confidence in the process.
- If you are collecting data in Excel but intend to analyze with Minitab, record your data in a column and it can then be copied and pasted into Minitab.
- 00:04 Hi, I'm Ray Sheen.
- 00:06 Now let's move into the best practices for the actual conduct of the Gage R&R Study.
- 00:11 And while I'm at it, I'll introduce the targets that we have for our analysis.
- 00:17 I'll start with some best practices for the measurement and
- 00:20 data collection portion of the study.
- 00:23 The challenge is to reduce the effect of noise that could create special
- 00:26 cause variation.
- 00:27 So we will periodically just mix things up a little bit to keep it as random as
- 00:31 possible.
- 00:32 Have one of the appraisers measure all the items one time in a random sequence.
- 00:36 Then have the next appraiser measure all the items one time in a different
- 00:40 random sequence.
- 00:41 Continue that pattern until all appraisers have measured everything once.
- 00:46 This is the first pass, or the first trial.
- 00:49 Now repeat the process for
- 00:50 the second trial, but be sure that you use a different random sequence.
- 00:54 And if there are three trials, do it yet one more time.
- 00:58 While the measurements are being taken, record the results on the form or
- 01:02 in the computer.
- 01:03 Be sure that the measurement is recorded with the correct item.
- 01:06 Even though they are measured in random order,
- 01:09 the data must be recorded with the proper item.
- 01:12 So now let's switch over and look at some best practices for the analysis.
- 01:17 The first analysis needs to actually be occurring while you're collecting data.
- 01:22 You need to assess the process to make sure everyone understands
- 01:25 what they're doing and they're following the plan.
- 01:28 I'd also carefully check to ensure that they are completing the data form
- 01:31 correctly.
- 01:32 Of course, once you have the data, you'll then do the appropriate analysis,
- 01:36 either variable analysis or attribute data analysis.
- 01:39 I'll go through those in detail on the next few lessons and
- 01:42 we'll work through an example.
- 01:44 Once the calculations are complete,
- 01:46 you need to assess whether the measurement system is acceptable for the application.
- 01:50 Now I'll go over the levels of acceptability and
- 01:52 marginal performance in the next few slides.
- 01:54 If the variation is too large, you need to either take actions to improve
- 01:58 the existing measurement system or switch to a different measurement system.
- 02:02 Either way, you'll need to do another Gage R&R study.
- 02:06 Let's take a moment and discuss the target results for
- 02:09 a variable data Gage R&R study.
- 02:12 A variable data study gives a single comprehensive result that
- 02:15 combines both repeatability and reproducibility.
- 02:18 The metric generated is known as GRR.
- 02:21 Because of the richness of the variable data,
- 02:24 we can get a very good answer with only a few items in the study.
- 02:28 While the GRR number is nice to have,
- 02:30 what really matters is the relationship of GRR to the normal part variation.
- 02:34 So we normally will express the GRR as a percentage of the total variation
- 02:38 within the study.
- 02:40 The target we will use is a maximum value of 10%.
- 02:43 That means that the common cause variation of measurement system error as captured by
- 02:48 the Gage RNR study contributes less than 10% to the measurement variation.
- 02:53 So what if we're at 11% or 12%?
- 02:56 Well, a GRR percentage of between 10% and
- 02:58 30% is considered to be marginal performance.
- 03:01 That means the system will detect big differences in the item values,
- 03:05 but probably not detect small differences.
- 03:08 Depending on the application, that may be all right.
- 03:11 If the GRR is over 30%, the system cannot be trusted.
- 03:14 When that is the case, you need to either change or improve the measurement system.
- 03:20 Now let's look at the targets for an attribute data Gage R&R study.
- 03:24 One of the big differences between attribute and
- 03:27 variable Gage R&R studies is that the attribute study gives separate values for
- 03:31 repeatability and reproducibility.
- 03:33 Whereas variable studies have one comprehensive measure,
- 03:36 although you can separate that out.
- 03:38 In addition, most attribute studies include references for
- 03:41 each item of its true quality status, either good or bad.
- 03:46 And with this reference, the study could evaluate both accuracy and
- 03:49 precision error.
- 03:50 This accuracy assessment is normally called effectiveness with these studies.
- 03:55 The analysis can be completed through a set of basic formulas
- 03:58 to get a pretty good approximate answer.
- 04:00 The statistical software uses some advanced statistical analysis based upon
- 04:04 the Chi-square statistical technique to get an even better answer.
- 04:08 It includes an additional measure of system performance that is known as Kappa.
- 04:12 However, the measures we have will normally answer our questions and
- 04:16 the differences between the two approaches are usually quite small.
- 04:19 The study does not compute error but rather computes success.
- 04:24 So we find that our targets for repeatability and
- 04:26 reproducibility are greater than 90%.
- 04:29 The marginal range is 80% to 90%, and anything less than 80% is unacceptable.
- 04:36 As I said, we will have some different measurements
- 04:38 which will include the reference standard in the analysis,
- 04:42 which means we can also calculate miss rate, false alarm rate and effectiveness.
- 04:47 The target for miss rate is 2%, false alarm rate 5% and
- 04:51 effectiveness is also 90%.
- 04:53 And the unacceptable ranges for those are at the 5%, 10% and 80% levels.
- 05:00 Well, we've looked at some best practices for running this study and
- 05:03 we've looked at the targets for performance.
- 05:06 All that's left is the calculations, which we'll cover in the remaining lessons.
Lesson notes are only available for subscribers.
PMI, PMP, CAPM and PMBOK are registered marks of the Project Management Institute, Inc.