Test for normality - non-normal data ?

My histograms shows the data are rightly skewed.

So if the data are not normally distributed, i can't calculate the control limits or use the "tests for special causes", does it mean I have to convert non-normal data to normal ?

if so, how do I convert

So after conversion, how do I proceed from there? Will the data be "transformed" and I have to use the new figures?

Do I need to do "sub-grouping" ? If so, how would you suggest I group the data so that I can do Variable Charts for Subgroups > Xbar-S (in minitab) ?

Thanks !

Re: Test for normality - non-normal data ?

Hey hazel.

You might want to tell us what you are trying to do. Are you trying to do some standard test like estimate a mean (or do an ANOVA)? Maybe look at residuals?

There are a couple of test-statistics that you can use to check normality: Things like Shapiro-Wilk and Kolmorogov-Smirnov help test for normality, but you should also try and find a function to create a "fitted" normal model of your data which should show a curve along your histogram.

Combine these two things together will help you get an idea both in graphic and test-statistic and p-value form whether something may be normally distributed (Remember that things can come from some distribution but they won't necessarily reflect the entirety of that distribution: we usually have enough confidence if the sample size is large enough, but even this is not any gaurantee).