I have a basic question - i understand while using I-MR chart the data has to be normal. my question is why must the data be normal.

 

 

 

Dear Jerry,

To my mind control charts do not need any particular data distribution. Aim is to find out signals i.e. presence of special cause of variation that needs to be investigated.

Regards.

K P Kapoor

thanks kanwal that sounds logical.

but it seems there is another school of thought which says that; its is possible for data to be non-normal for a process & the process in reality is perfectly fine. in this case IMR will show process is out of control & the process has to be tweaked - to be in control which in reality might not be required

appreciate ur inputs on the above.

 

Dear Jerry,

The control limits (UCL & LCL) are based on +/- three sigma limits away from the center (average). These control limits are used to determine if a process is in control or not.

Only if the data is normally distributed will 99.73% of the data lie within the three sigma limits. So, we expect that most of the data will fall within the control limits. If data points don't fall here, then in all probability, the process has gone out of control.

For other types of data distributions, the % of points that lie within the three sigma limits will be different. Let's say only 80% of the data lies within the three sigma limits. In which case, even if the process is in control, the control limits will indicate a lot of false positives where it will show that the process is not in control even if it is in control.

Other thoughts on this question?

SJ>

To add to this, in a skewed distribution, 3 sigma from mean may cover most data on one side and may not cover enough on the other side. So as SJ said, chances of false signals exist.

However, IMR is used for attribute data (which is non-normal) as well by some companies for symmetric distributions (not skewed) and several authors support it by saying that it works reasonably well. 

Thank you Dr. Suresh & Vishwadeep Sir...........this has clarified my doubt. I really appreciate ur time & inputs.

Regards,

Jerry