Control Charts for 29 Devices that I tested

[jdm2008]

I have 29 devices that I tested. I am trying to generate a control chart for this study. These devices take a measurement. When testing them, I know for certain the measurement in reality is zero. So at two points I read the devices and see what they indicate. These are after 3 minutes and 15 minutes.
I am fairly certain that in this case that “n” would be 29 and this number is what all of the constants I would need are. When I put the data into a control chart calculator I’ve discovered the constants for n=2. Am I correct?

 

[Jim Wynne]

I have 29 devices that I tested. I am trying to generate a control chart for this study. These devices take a measurement. When testing them, I know for certain the measurement in reality is zero. So at two points I read the devices and see what they indicate. These are after 3 minutes and 15 minutes.
I am fairly certain that in this case that “n” would be 29 and this number is what all of the constants I would need are. When I put the data into a control chart calculator I’ve discovered the constants for n=2. Am I correct?

What are you trying to find out? What sort of "control chart calculator" are you using? What kind of chart do you want to produce? Are you charting the 3-minute data and the 15-minute data separately? Are the data in chronological order?

 

[Bev D]

Jim is correct: let's take a step back and understand the question you are trying to answer. It doesn't sound like a 'control chart' is the analytical tool you need to answer your real question.

 

[jdm2008]

I want(I've been instructed) to produce a control chart. The control chart calculator just gives the the control limits.
The 3 minute data and 15 minute data are not charted seperately.

 

[Jim Wynne]

I want(I've been instructed) to produce a control chart. The control chart calculator just gives the the control limits.
The 3 minute data and 15 minute data are not charted seperately.

At this point all we know is that you want to produce a control chart and you're using a calculator of some sort. Without knowing (a) what you're trying to learn about the data, (b) the nature of the "calculator" and the underlying assumptions of the software and (c) why you have two data sets that you're lumping together, there really isn't much we can say that would be helpful. Please provide as much detail as you can.

 

[jdm2008]

Thanks for the reply I was debating whether to include more I guess i should have.

I have 29 devices that measure a quantity. I set the quantity measure independently of each device at zero. There are reports that the devices measure a quantity other than 0 when they should measure 0. This defect supposedly happens during the begining of its operation(ie if the device sits on for 30 minutes and does nothing this problem disapears). The data sets are lumped together because there should be no change from data taken at 3 minutes versus data taken at 15.
However there may be variation. They should both be 0(or some number close to 0), we are testing this point. I am trying to learn how close the numbers are too zero, and whether they are different outside of normal random variation.

As to the calculator, I'm not so much concerned about the calculator because I can do the control limits by hand. I'm concerned about whether in this case D4 and A2 would be for n=29 versus n=2.

 

[Jim Wynne]

Thanks for the reply I was debating whether to include more I guess i should have.

I have 29 devices that measure a quantity. I set the quantity measure independently of each device at zero. There are reports that the devices measure a quantity other than 0 when they should measure 0. This defect supposedly happens during the begining of its operation(ie if the device sits on for 30 minutes and does nothing this problem disapears). The data sets are lumped together because there should be no change from data taken at 3 minutes versus data taken at 15.
However there may be variation. They should both be 0(or some number close to 0), we are testing this point. I am trying to learn how close the numbers are too zero, and whether they are different outside of normal random variation.

As to the calculator, I'm not so much concerned about the calculator because I can do the control limits by hand. I'm concerned about whether in this case D4 and A2 would be for n=29 versus n=2.

I'm sorry, but maybe I'm a little thick because I still don't have a clear picture of what's going on, although I think I'm a little closer now. Some of the things we still don't know:

• what's being measured;
• how it's being measured--it is a continuous stream of data into the devices, or periodic measurements, e.g.;
• why you've gathered two sets of data (3-minute & 15-minute) if you're not concerned about differentiating between them in your analysis;
• why you're apparently trying to do calculations for subgrouped data (hence your questions about A2 and D4) when there don't appear to be rational subgroups involved;
• if you're trying to find out if the devices are producing spurious readings--giving results >0 when you think that the input value should be 0, or if you're trying to find out if the 29 devices produce acceptably similar results, or what.

It still appears that a control chart isn't appropriate here, but if it is, it's bound to be an I/MR chart you're looking for. Have a look through this thread and see if there's any help there.

 

[jdm2008]

-specifically nuclear activity is being measured inside an enclosed space. Since the source of nuclear activity isn't present we know the reading should be 0 or close to i.
-A reading off the device. At a particular point.
-In theory the readings should be the same. We have a suspicion(or more correctly a suspicion was relayed) that they are not however.
-I would say acceptably similiar results both being 0.

 

[Jim Wynne]

-specifically nuclear activity is being measured inside an enclosed space. Since the source of nuclear activity isn't present we know the reading should be 0 or close to i.
-A reading off the device. At a particular point.
-In theory the readings should be the same. We have a suspicion(or more correctly a suspicion was relayed) that they are not however.
-I would say acceptably similiar results both being 0.

For purposes of clarity I've combined my questions (Q) and your answers (A) and added my responses in italics (R):

Q: What's being measured?
A: Specifically nuclear activity is being measured inside an enclosed space. Since the source of nuclear activity isn't present we know the reading should be 0 or close to it.

R: This calls the function of the measurement devices into question; if you're dealing with radiation (what I assume you mean by "nuclear activity) detection, why are the devices registering detection of the target phenomenon if you know that it isn't present? I this what you're trying to find out?

Q: How is it being measured--it is a continuous stream of data into the devices, or periodic measurements, e.g.?
A: A reading off the device. At a particular point.

R: So I assume now that it's a continuous data stream and readings were taken at intervals.

Q: Why have you gathered two sets of data (3-minute & 15-minute) if you're not concerned about differentiating between them in your analysis?
A: In theory the readings should be the same. We have a suspicion(or more correctly a suspicion was relayed) that they are not however.

R: You have two data sets, and you want to compare them? To what end? This is still not clear to me. Are you trying to chart the two sets individually (this goes back to your original question regarding n=2 or n=29)?

Q: Why are you apparently trying to do calculations for subgrouped data (hence your questions about A2 and D4) when there don't appear to be rational subgroups involved?
A: Not answered

R: An answer to this question would be good.

Q: Are you trying to find out if the devices are producing spurious readings--giving results >0 when you think that the input value should be 0, or are you trying to find out if the 29 devices produce acceptably similar results, or what?
A: In theory the readings should be the same. We have a suspicion(or more correctly a suspicion was relayed) that they are not however.
-I would say acceptably similiar results both being 0.

R: It now seems certain that control charts of any variety are not going to help you.

Although I'm still not clear on the whole thing (such as why there are 29 measurement devices, a question I didn't ask), it seems to me that you need either some sort of controlled experimentation, or perhaps some form of measurement system analysis (MSA) or some combination of the two.

 

[Sturmkind]

I agree with Jim. This seems more like an attribute case of either the radioactivity is present and detectable or not. Readings close to zero probably represent the measurement 'noise' within the system and are likely to be considered as gage uncertainty. Otherwise, the light is either on/off and we would be uninterested in knowing if the light is 'almost' on or off.

With the data already in hand, then the device with the maximum range or value from zero should be known. The remaining question would be whether that device varies over time.