0

I have a sample of 30 data points, that I am unable to find a distribution fit for. The goal of my analysis is to assess process capability and get an accurate Ppk. This is one output of a process I am validating.

Unable to fit anything, I observed that assuming normality, the data gives a Ppk>3.00. The data set is tailed to one side due to one or two (true and valid) outliers. These outliers are representative of the process and I cannot modify and improve the process to remove these and then re-sample later.

So I'm trying to find out what my options would be. Re-sampling and measuring again would likely yield similar results and is time consuming/laborious. Is there some rationale I can use to allow me to assume the data normal for analysis? Would be your next move in a similar situation?

Really appreciative of any help here, I'm stumped!

Thanks tom

tombyrne412
  • 1
  • Hi and welcome to CV. It might help if you provide more information on what you are trying to achieve. Do you want to approximate an experimental distribution? And if so, what is the purpose of the approximation? – cherub Apr 30 '19 at 07:44
  • @cherub I need to calculate process capability (Ppk, Pp). However I am unable to fit the data to a known distribution. Essentially, I am looking for information around distribution-free analyses, or what else would be a good avenue to try. – tombyrne412 Apr 30 '19 at 07:58
  • 1
    What is Ppk, Pp? – Gijs Apr 30 '19 at 08:35
  • please explain further the nature of your data . how did you record it ? – IrishStat Apr 30 '19 at 09:19
  • @IrishStat The data was sampled randomly from a large (N=750) population. Recorded by hand using a previously validated test method for a very similar output (similar enough to be rationaled equiv.) – tombyrne412 Apr 30 '19 at 10:27
  • @Gijs apologies, when I say I am assessing process capabiltiy, I mean to say I am assessing the process performance. That is why I am looking for accurate Ppk and Pp indices. – tombyrne412 Apr 30 '19 at 10:30
  • Hi again, I'm (also) still puzzled by your question. The abbreviations and therefore the meaning of Ppk and Pp are not familiar. Are you looking for a way to get a tight(er) grip on the statistics, or on the interpretation of your analysis? If you have some data and are trying to describe this, you are in well known statistics territory. If there are special/specific methods and outcomes, probably only someone from your field will be able to answer. – cherub Apr 30 '19 at 13:46

0 Answers0