Variable data w/ varying sample size
Six Sigma – iSixSigma › Forums › Old Forums › General › Variable data w/ varying sample size
 This topic has 7 replies, 4 voices, and was last updated 16 years, 6 months ago by ALEK DE.

AuthorPosts

May 9, 2005 at 5:26 pm #39278
What is the best chart type to construct given variable data and a varying sample size? The mean for a service process is calculated weekly with a population that varies roughly between 40 – 60. Given the varying sample size, what is the best way to calculate an upper control limit for this process? Can A and B values be used with a mean with sample sizes this large, and if so how do I use A and B values with varying sample sizes. Thank you in advance.
0May 9, 2005 at 6:10 pm #119197Subgroup the data daily (if it makes sense), average it, and plot it on an xbar/R if n is between 29 or xbar/s if greater thatn 10. Use excel or minitab for UCL and LCL calculation.
0May 9, 2005 at 7:34 pm #119203Individuals & moving range chart if you want one set of control limits for all the data. The method mentioned by Utah calculates different control limits for each subgroup.
0May 11, 2005 at 6:28 pm #119347Does it make sense for me to use the Xbar/s chart given a population mean or is this only valid for samples? Thanks
0May 11, 2005 at 7:18 pm #119352The statisticians out there will probably smell blood in the water, but I almost always use the range method for estimating variation, just because it is easier to understand and compute than s. There are (strong) arguments that show the error in the variation component using R charts are significant for n>10 {or >12 (Juran) or >15 (Wheeler)}, but I rarely use subgroups this large. For the individuals chart, n=1.
If you’re using MINITAB, the “help” icon on each command does a good job of describing the use of each chart and also shows the equations used. Might try reading these and decide which chart makes more sense to you. Look at the IMR versus the IMRR/S options.
Hope this helps.0May 26, 2005 at 8:00 am #120225
ALEK DEParticipant@ALEKDE Include @ALEKDE in your post and this person will
be notified via email.FTSBB, A small correction. In IMR , n is generally 2.
0May 26, 2005 at 12:47 pm #120245For individuals chart n=1, hence the name “individual”. The bias correction factors and control constants use the values for n=2, as you are comparing two subsequent measures. See Understanding Statistical Process Control by Donald Wheeler, page 48, for a good example. There are probably countless other texts on IMR – take your pick.
0May 26, 2005 at 2:27 pm #120260
ALEK DEParticipant@ALEKDE Include @ALEKDE in your post and this person will
be notified via email.Possibly it’s a different way to understand thing. My interpretation of a group is no of samples in a sub group & that’s important when we use Range as estimate of variation , d2 constatnt changes with the the no of samples in a sub group. In I/MR , moving range is generally with two subsequent data points & hence I commented ( becaz here d2 is corresponding to sample size2). In individual chart it’s always individual – no confusion on that. Hope we are talking the same thing.
Thanks.0 
AuthorPosts
The forum ‘General’ is closed to new topics and replies.