Posted: 10/3/2016 5:55:59 PM EDT
|
Finishing up a stats class. Have a bonus question I'd like to run past some folks with more knowledge than me. I've tried Google and a few other things but none of it makes sense in my head. (We've been working on run charts and histograms and finding upper and lower control limits after identifying variance and standard deviation,
Here's the question: You are working in a Process Improvement Team on a process that turns out to be unstable because one of it's data points is below the lower control limit. The team studies the process further and is able to eliminate the cause of the outlier. The resulting process is stable but producing results outside the specification limits due to a high standard deviation. What do you recommend the team should do next? Any thoughts our direction would be appreciated. |
|
The resulting process is stable but producing results outside the specification limits due to a high standard deviation. If, after removing the outlier legitimately, the process is stable but out of tolerance conditions exist, then the Cp value (spread of the data) is good but not the Cpk...you need to adjust the mean to a more nominal condition. ETA: the "high Std Dev" issues bugs me though. I'm not convinced you have a 1.33 Cpk with a "high stddev"...let me kick some numbers around. |
|
Quoted:
Fudge the numbers to get it within specification limits. This. Meaning convert the numbers into usable data. I always thought it was for when the numbers where all over the place, but it might work. or run another test/use different samples? Run different test? Are you sure they used the right one? |
|
Quoted: but how? I think that's what they are trying to do. Quoted: Quoted: Too much standard deviation means the process has too low a Cp value. Improve the process. but how? I think that's what they are trying to do. |
|
Quoted:
but how? I think that's what they are trying to do. Quoted:
Quoted:
Too much standard deviation means the process has too low a Cp value. Improve the process. but how? I think that's what they are trying to do. Fuck it...go X-bar, bar...average of the average. 
Group the data into subgroups and run that. There are lies, damn lies, and then; statistics. |
|
Quoted:
Fuck it...go X-bar, bar...average of the average.
Group the data into subgroups and run that. There are lies, damn lies, and then; statistics. Quoted:
Quoted:
Quoted:
Too much standard deviation means the process has too low a Cp value. Improve the process. but how? I think that's what they are trying to do. Fuck it...go X-bar, bar...average of the average.
Group the data into subgroups and run that. There are lies, damn lies, and then; statistics. I like The Law of Lesser and Greater Errors. It all balances out in the end. |