Is this Data a Normal Distribution?

[Booker]

Should I expect this set of data as normal distribution. And can someone simply clarify hot to determine if the data are normal distributed or not.

I am using MINITAB.

Thanks

 

[Miner]

Re: Is this Normal Distribution?

This distribution fails the normal probability plot (middle right p = 0.018 < 0.05). However, this is probably due to the fact that your process is unstable (upper left, first 8 points above the mean). If your process were in a state of control, the resulting distribution would most likely be normally distributed.

Your first priority should be controlling the process.

 

[sylwia]

P value is sometimes not good for normality test.
To be sure, that your test about normaity is correct, please check skewness and kurtosis--> they must be for normal distribution: -0,5 to 0,5 (the best: both= 0); in minitab: Stat- basic statistic- Graphical summary

 

[Booker]

P value is sometimes not good for normality test.

Thanks for hint, but what does mean "sometimes". When P value is below 0,05?

 

[bobdoering]

Thanks for hint, but what does mean "sometimes". When P value is below 0,05?

I would recommend reading this concerning the issue of p value and normality. My preference is to do a best curve fit analysis.

 

[Bev D]

Re: Is this Normal Distribution?

?..However, this is probably due to the fact that your process is unstable (upper left, first 8 points above the mean). If your process were in a state of control, the resulting distribution would most likely be normally distributed.

Your first priority should be controlling the process.

To clarify: If THIS process were in statistical control it would most likely be Normal. But this is not a universal truth. Processes can have any distribution and still be stable and predictable.

Not sure if the statistical out of control condition is of any practical importance given the distribution of the process to the specification limits. This may be a simple non homogenous process that is very tightly controlled. We would need more data and more information on the process to determine that...

 

[bobdoering]

Re: Is this Normal Distribution?

To clarify: If THIS process were in statistical control it would most likely be Normal. But this is not a universal truth. Processes can have any distribution and still be stable and predictable.

Even if the data was normal, that is not a guarantee the process is. What is the amount of contributing gage and measurement error? They have the ability to make any process appear normal if they mask the underlying process distribution. They tend to be normal, and fake out most people.

Does this variation represent what you expect to see across the life of the process? Does it always start up at the same level? Maybe short term variation is not the issue.

I know, myopic customers "just want a number" to check of the capability box.

 

[Apoch]

As bobdoering has stated in his post, the measurement system should be first assessed whether it is acceptable. That should be the first premise in the data analysis.

Provided that the measurement system is deemed to be adequate, a test for normality would then ensue. As others have mentioned, the p-value from the p-plot can be used to assessed whether the data follows a normal distribution. A p-value < 0.05 indicates a poor fit to the normal curve.

Normality testing is very important, the Cpk being reflected by Minitab assumes a normal distribution. The index, however, would then result in a misleading analysis of how your process is performing.

 

[Booker]

My question is general, how to recognize Normal Distribution.
Anyway, these data are real from one of our supplier during PPAPing. If I saw this I requested new capability study, but after that when new MSA will be performed, with "some" perfect result.

 

[bobdoering]

My question is general, how to recognize Normal Distribution.

My response can be found here, sine I have used the response many times before.

Anyway, these data are real from one of our supplier during PPAPing. If I saw this I requested new capability study, but after that when new MSA will be performed, with "some" perfect result.

The real question, as Bev D indicated, is should it be "normal"?