Data normality

[nickfenton13]

Hi all,

I have been asked to do a capability study in regards to tensile strength results. My problem is, the data does not seem normal and after constructing a Probability
plot, it seems to confirm this. I have attached two photos of the probability plots created. Is the data close enough to normal to be able to then conduct a capability study or does it require transformation of some sort? Also, why does the data seem much closer to the normal when taking z scores instead of CDF?

My other question is; is it worth doing a capability study on a one sided specification or is there a better form of data analysis I could use to analyse this kind of data?

 

[Romsek]

You don't give us much to go on. What is your analysis trying to show?

You can make conclusions about things w/o your underlying properties being normal.

Also those plots look pretty normal to me. Have you done any of the typical normality tests?

Are the odd moments of the normalized data zero?

 

[Deleted member 4993]

Fracture strength data generally follow log-normal or Weibull distribution.