Home > Statistics Every Writer Should Know > The Stats Board > Discusssion
Non normal distributions
Message posted by carlos pe鎙 (via 194.30.71.242) on February 18, 2002 at 5:00 AM (ET)
When you have a non normal distribution (e.g. concentricity, planarity,..), what are the best methods for:
- use the control graphs and calculate the limits?
- calculate the capacity of the process?
Could be precontrol an option?
Thank you
READERS RESPOND:
(In chronological order. Most recent at the bottom.)
Re: Non normal distributions
Message posted by Darius (via 200.23.217.10) on February 18, 2002 at 10:18 AM (ET)
For a non normal distribution:In control charts and capacity studies the best option is non parametrics. The use of percentil insteed of the traditional methods.
Could be precontrol an option?
Most of statisticians hate (literally) PreControl, but I tink is a good approach if:
* If you can愒 control your process (it愀 not in your hands to control)
* If to accomplish the specifications is more important than reduce variation.
* If there is no effect in the quality of your products while variation is inside the specification limits.
But not use precontrol because of non-normality
Re: Non normal distributions
Message posted by Darius (via 200.23.217.10) on February 18, 2002 at 10:35 AM (ET)
But as Donald Wheeler said, "Control charts work good even when non normality occur (because, there is not going to be false alarms)".Your problem may be autocorrelated data.
Re: Non normal distributions
Message posted by Jack Tomsky (via 12.144.103.66) on February 18, 2002 at 1:30 PM (ET)
Usually the most effective way of dealing with non-normal distributions for control charts is to find the simplest transformation which makes the data normal. Often the original units are not linear (e.g., areas, volumes, weights) and a transformation to linear units can make them near-normal.
Re: Non normal distributions
Message posted by JG (via 128.8.22.154) on February 18, 2002 at 7:54 PM (ET)
You should also consider averaging and the Central Limit theorom.
Your $5 contribution helps cover part the $500 annual cost of keeping this site online.
