# SPC and Normality test

Six Sigma – iSixSigma Forums Old Forums General SPC and Normality test

Viewing 5 posts - 1 through 5 (of 5 total)
• Author
Posts
• #37333

ROSS
Member

Hi:
Does normality test is must when we perform SPC for continous value? I remember it is no need for  “The central limit theorem”.
Tks!

0
#109696

Ken Feldman
Participant

Most continuous control charts (xbar and R,s) are robust to normality.  No, the Central Limit Theorem does not really apply to the control charts.  Don Wheeler has written some nice explanations of this.

0
#109699

ROSS
Member

Hi Darth:
Thanks for your instruction very much! Many warmhearted experts here and that is why so many people here. And I saw “Forums Say Goodbye to Stan and Darth”,  pls forget it and just continuous, or this forum will die offf just as aBBinMN said.
And do you mean we can perform control charts (xbar and R,s) for continuous value even they are not normality distribution? One of my friend asked me how to deal SPC for nonnormality distribution.
Tks!

0
#109700

Ken Feldman
Participant

Thanks for the kind comments.  Here is a good explanation I found on the Web.
Is normality of the original distribution a requirement to use control charts?  Depending on what literature you read the answer to this question will vary.  Wheeler (1992) points out that the assumption of normally distributed data is introduced into the control chart formulas through the use of control chart constants (used to compute the control limits).  The values for these constants are computed using a normal probability distribution for the original data.  However, Wheeler (1992) states that the control chart constants do not change appreciably even when the data display a non-normal distribution.   In fact, he demonstrated that control charts work well for non-normal data through the use of simulation studies of several skewed theoretical distributions.  Therefore, the data do not have to be normally distributed before using control charts.  The normality of the data is neither a prerequisite nor a consequence of a state of statistical control (Wheeler, 1992).

0
#112412

Neeraj Singh
Participant

Hi,

Dear Bob/Mike……. can you help on this………..I am a little cofused here. Could somebody tell me that if it really doesnot matter about the data being normal or non-normal for the control charts of continuous data say for eg. for I & MR chart, then why do we have different formulae to deal with this problem. The best eg. would be about the centre line of the control chart itself. In case of normal data, we compute the mean of the continuous data for centre line ( I & MR)whereas in case of a non-normal data we use the median as the centre line( I & MR). similarly, the UCL & LCL too have different calculations…Bob, it would be very nice if youcould mail me too on the same………awaiting the same….Rgds/*-..NK

0
Viewing 5 posts - 1 through 5 (of 5 total)

The forum ‘General’ is closed to new topics and replies.