INITIAL POSTING: believe this is a simple question for most of you, but, I am having trouble as a rookie: I have a truck of incoming foodstuff and I want to find out minimum sample size to take samples that will represent the each truck's moisture. Each truck approx. weighs 20 mt of foodstuff. How much kg of sample should I take as a min?
RESPONSE: Your question is a very good question. Often questions are asked as a sample size question; however, something else is really needed.
I will make some assumptions about your situation. First, I will assume that your customer is interested in the moisture content of a portion of your tank car, as opposed to an overall average; i.e., I will assume that there is variability of moisture within the tank car and that this is important relative to how your customer uses the product.
I will also assume that you really also would like to REALLY make a statement about the overall process and make a future prediction; i.e., not just make a pass/fail decision about an individual tank car.
Here is a basic strategy that I would recommend to accomplish this; however, physical situations, etc. would lead to adjustments of the procedures:
1. For each tank car take 10 samples at 10 different "random" locations within the tank car. I chose 10 for a particular reason for this example; however, a smaller number could be used with a slight motification to the procedure below.
2. The mean of moisture of tank cars over time would then be plotted on an INDIVIDUALS chart.
3. The within-tank standard deviation of moisture within tank cars over time would then be plotted on an INDIVIDUALS control chart.
4. If both control charts are in-control, we say that we have a predictable process; i.e., are statements are not only about the past but also the future, unless something changes.
5. In Minitab conduct a capability analysis (stat>quality tools>capability analysis). Enter your specification. The 10 samples within the tank car will be "Std dev within", while the "std dev overall" considers within and between tank car variability.
6. The only think that I look at on this output is PPM total, which describes the overall percentage of samples that you take which you estimate to be out of specification (not only the past but the future -- unless something changes).
7. If this percentage is not satisfactory, you need to change the overall process. That is you are pulling for a Lean Six Sigma project.
Note that this strategy is very different than the typical strategy of sampling a tank car and trying to determine if the tank car is OK. This generic approach can lead to fire fighting.
Community metadata Could not be loaded. No pre-configured community forum parent found for the incoming fid: . You need to add this forum into custom_code/community.sql
Send to friend
Comments
forrestbreyfogle 3/17/2005
INITIAL POSTING: believe this is a simple question for most of you, but, I am having trouble as a rookie: I have a truck of incoming foodstuff and I want to find out minimum sample size to take samples that will represent the each truck's moisture. Each truck approx. weighs 20 mt of foodstuff. How much kg of sample should I take as a min?
RESPONSE: Your question is a very good question. Often questions are asked as a sample size question; however, something else is really needed.
I will make some assumptions about your situation. First, I will assume that your customer is interested in the moisture content of a portion of your tank car, as opposed to an overall average; i.e., I will assume that there is variability of moisture within the tank car and that this is important relative to how your customer uses the product.
I will also assume that you really also would like to REALLY make a statement about the overall process and make a future prediction; i.e., not just make a pass/fail decision about an individual tank car.
Here is a basic strategy that I would recommend to accomplish this; however, physical situations, etc. would lead to adjustments of the procedures:
1. For each tank car take 10 samples at 10 different "random" locations within the tank car. I chose 10 for a particular reason for this example; however, a smaller number could be used with a slight motification to the procedure below.
2. The mean of moisture of tank cars over time would then be plotted on an INDIVIDUALS chart.
3. The within-tank standard deviation of moisture within tank cars over time would then be plotted on an INDIVIDUALS control chart.
4. If both control charts are in-control, we say that we have a predictable process; i.e., are statements are not only about the past but also the future, unless something changes.
5. In Minitab conduct a capability analysis (stat>quality tools>capability analysis). Enter your specification. The 10 samples within the tank car will be "Std dev within", while the "std dev overall" considers within and between tank car variability.
6. The only think that I look at on this output is PPM total, which describes the overall percentage of samples that you take which you estimate to be out of specification (not only the past but the future -- unless something changes).
7. If this percentage is not satisfactory, you need to change the overall process. That is you are pulling for a Lean Six Sigma project.
Note that this strategy is very different than the typical strategy of sampling a tank car and trying to determine if the tank car is OK. This generic approach can lead to fire fighting.
Let me know if you would like to discuss further.
Forrest Breyfogle
Six Sigma Moderator for Quality Digest
512-918-0280 x401
www.smartersolutions.com
forrest@smartersolutions.com