95% confidence interval, or is it?

[#1]

Hello,

I have been asked to provide some statistical advice on process related values. So we are making a product and we are monitoring the amount of solvent left at the end of synthesis. 

I have been given 12 values. From this admitedly very small statistical sample, we want to make a prediction to say: in the future 95% of the values will be under this value. I tried giving my colleagues the upper 95% confidence interval from the distribution platform but they said it was not the right thing. 
One of my colleagues mentionned it may be necessary to generate predicted values in JMP, and if possible take into account the uncertainty of the measurement on these values. 

Any advice?

values below:

9,1
9,6
9,5
9,34
9,54
9,73
9,79
9,85
9,9
9,95
9,86
9,51

[#2 · Accepted solution]

The 95% confidence interval quoted in the Distribution platform is for the population mean...not individual values. It sounds like you are trying to estimate a tolerance interval on future values. The desire to predict future values means you have to make an assumption that the sample you have is reasonable and representative of the population from a statistical and practical perspective. Once you make that intellectual leap...then just use JMP's tolerance interval capabilities to estimate the quantities desired. Here is a link to the JMP documentation for tolerance intervals. Tolerance intervals in JMP 

[#3]

Hi @P_Bartell ,

thanks for your reply. So I have made the leap of assuming this data is a representative sample. so I used the  confidence interval on the distribution. What I'm not sure about is which value to use.

As I want to predict the limit under which 95% of values will be, I've selected unilateral upper limit and 0.95 as the proportion to cover. But what about the confidence level? should I just keep it at 0.95?  

[#4]

You can set the confidence level any place you like. What is more important is that when you communicate the tolerance limit you in the same breath communicate the confidence level used as part of determining the tolerance limit.