Which control chart to use?
Six Sigma – iSixSigma › Forums › Old Forums › General › Which control chart to use?
 This topic has 12 replies, 9 voices, and was last updated 17 years, 2 months ago by Cesar Montoya.

AuthorPosts

September 9, 2002 at 1:10 pm #30295
I am currently looking to control chart my monthly data, up until now I have been using Excel to chart my data. I have continuous data and a population size between 190 & 300 each month. I get the data all together from a database. It does not make sense to pull a sample from this but to use the whole population.
Which control chart would be best?
I was thinking about x bar & s, but it is designed for a consistent sample size.
Please help!!
0September 9, 2002 at 1:27 pm #78731
James AParticipant@JamesA Include @JamesA in your post and this person will
be notified via email.Marnie,
If you already have an historical value for s then I would be tempted to use X bar moving R charts (or soft equivalent with a sample size of 1) that way you can use all the points from the previous month, and use the existing svalue (+/ 3sd for limits), if there has been no change to the process. The only question I have is – What will you react to 30 days after it happened? It must be a fairly stable process, so do you actually need to run SPC on it? Or is it something like absenteeism and just for the record?
Intrigued. Anyone else got other suggestions?
James A
0September 9, 2002 at 1:46 pm #78734James,
The process is fairly stable. I am at the point of closing the project off and need the control chart for continual monitoring (company requirement). The project is on waste, the data I am trying to charts is waste per customer order, while the team looks at waste created per input (daily average). We do not look at this data more than once a month, this is to ensure that improvements are working. Thanks for the help!
Marnie0September 9, 2002 at 2:25 pm #78738
James AParticipant@JamesA Include @JamesA in your post and this person will
be notified via email.My pleasure. Thanks for the explanation, it makes sense, now.
Smile, it’s Monday!!!!
James A0September 9, 2002 at 6:08 pm #78754Marnie,
Actually, the XbarS chart would be the appropriate chart for varying sample size. The XbarR chart has a requirement of consistent sample size. I don’t know your process, nor the coomplexity/time to pull and analyze the data, but there should come a time where we feel confident in pulling representative sampling. We shouldn’t feel the need to pull all available information to estimate our Voice of the Process. With this accomplished, rational subgrouping becomes a critical concept.0September 9, 2002 at 6:26 pm #78757
Just meParticipant@Justme Include @Justme in your post and this person will
be notified via email.If you are going to use a full months data as a subgroup, I asume you are going to have one representing data point per month for the mean and one for the stanadard deviation? A control chart’s primary function is to identify an out of control condition as early on as possible so the appropriate personnel can respond accordingly. Are you not concerned about having month to month feedback? Do you feel the information might be to little too late? In my experience it is best to develop control charts to give me early warnings of potential issues? I am probably just unclear on your application?
Best wishes to you0September 9, 2002 at 6:26 pm #78758Erik,
As I read James note over again, I agree with you. The control chart I used was the Xbar & S (which is what I thought he ment by the Xbar & moving R – but I see he was talking about the I&MR chart). The data is more easily obtained as a population than as a sample. I have little confidence in sampling from this population due to the amount of variation in the inidividual data points. The variation between the monthly averages is much less then the individuals. I would agree with sampling when the data is normal and the variation is acceptable. For this process moving the target is more important than reducing the variation in the individual data.
Thanks for the input.0September 9, 2002 at 6:39 pm #78759Just me…
I am using one data point to monitor the mean and one the std dev. I am not concerned about month to month feedback. This control chart is part of the control plan (postproject hand off). To monitor the data any more would tie up too much of my time and prevent me from taking on new projects. The data has a HUGE amount of variation between customer orders due to the way in which we manage our customer orders. It is too difficult & thus expensive to control the individual customer order waste (at this time). That is the reason for blending the information on a monthly average basis.
Thanks for the input. This is helping me prepare for my defense.0September 9, 2002 at 8:35 pm #78763
Robert ButlerParticipant@rbutler Include @rbutler in your post and this person will
be notified via email.Based on your posts it sounds as if your primary interest is in having some form of graphical representation of your data to give you some idea of where you have been and a sense of where you might be going rather than trying to control anything in particular. Given that you are taking your data in monthly blocks, a much better way to view the results of your process would be to run monthly boxplots and plot them over time.
Your posts lead me to understand that your data may not be normal and that the issue is really waste per customer order as opposed to average waste for some meaningful groupings of customer types. In this instance, monthly boxplots make even more sense. If you want to have some “control limits” for the boxplots take your prior data, plot it on normal probability paper and identify the .135% and 99.865% points and use these as your +3 standard deviations limits to visually assess how your process is doing over time.0September 9, 2002 at 9:11 pm #78766
GabrielParticipant@Gabriel Include @Gabriel in your post and this person will
be notified via email.Marnie:
If you just want to make a follow up of the trend in the average and variation, you can still usa a Xbar S chart, without control limits. After all, you are not trying to decide incontrol / outofcontrol.
If you still want control limits, you still can use any control chart (even XbarR or XbarS), provided taht you know the process average standard deviation (which is usually the case when you are making a follow up). You can calculate the control limits point by point.
For example, if using a XbarR chart, with the sample size of a given point you have d2, and with d2 and the already known process standard deviation you have Rbar = “sigma” x d2. With the sample size of the point you also have D3 and A2. And with these constants, together with tha already known average and the already calculated Rbar you have the control limits for this point.0September 10, 2002 at 12:13 pm #78774Charting is not something to take lightly, it is also not something that you just grab data and find a chart to drop it into.
1) What information do you wish to gather from the chart? If you want it to be an analytic tool that alerts you to a problem before it gets to large you must use sampling data carefully constructed to absorb the variable factors involved with the process. The premier tool for this is a XBar&R ( XBar &S if your sample sizes are very large 25 or greater).
2) Normality is not a consideration when constructing control charts the nature of the charts and the interpretive value is not affected by normality.
3) ImR charts can be utilized when single point data is the only thing available, however, their ability as analytic tools is very limited.
My suggestion is to rethink what it is you need from this process then design a sampling plan to accumulate the data and place it on a XBar&R chart.
0September 10, 2002 at 12:21 pm #78775
bob fareyParticipant@bobfarey Include @bobfarey in your post and this person will
be notified via email.In my opinion, you should plot, in real time, individual data values on an X, moving range chart. This will demonstrate whether your process is stable, and if so, what the mean and upper / lower control limits are. If it is not stable (drifting, cyclical patterns, points outside control limits etc), you can decide if you need to take action. You should be able to link your database to excel so that the chart plots in real time. Hope this helps.
0October 30, 2004 at 4:55 pm #110050
Cesar MontoyaParticipant@CesarMontoya Include @CesarMontoya in your post and this person will
be notified via email.To all:
How many data points do you need to create a good control chart?
Thank’s for your respond0 
AuthorPosts
The forum ‘General’ is closed to new topics and replies.