Sample size for correlation study of three form error measurement methods

Join the discussion
Registration is free. Ask a follow-up in this thread, or start your own.
Sign up Log in
4 replies · 2K views
Ranger Mike
Science Advisor
Messages
2,463
Reaction score
447
Hello all

First time on the Math side of this great Forum.

I need to do a correlation study.

There are 3 methods to measure Form Error of a round metal part.

  1. The base line is the traditional measuring method that is laborious but of the highest accuracy.
  2. A cnc machine measurement machine takes data with an analog tactile scan probe
  3. Lastly, a very sophisticated digiatal tactile scan probe.

Previous observations between the three methods have variacen of ½ to 1 micron.I am seeking a very strong correlation coefficient and high confidence level.

What would be a good sample size?

Thank you
 

Discussion

Physics news on Phys.org
Correlation coefficient between what, confidence level of what?

A proper correlation coefficient shouldn‘t be expected to depend on the sample size.
 
Ranger Mike said:
Hello all

First time on the Math side of this great Forum.

I need to do a correlation study.

There are 3 methods to measure Form Error of a round metal part.

  1. The base line is the traditional measuring method that is laborious but of the highest accuracy.
  2. A cnc machine measurement machine takes data with an analog tactile scan probe
  3. Lastly, a very sophisticated digiatal tactile scan probe.

Previous observations between the three methods have variacen of ½ to 1 micron.I am seeking a very strong correlation coefficient and high confidence level.

What would be a good sample size?

Thank you
If you want to test for the existence of a linear relationship, the standard statistic is the t statistic ( Df=n-2) for a sample of size ##n## and sample correlation ##r##, is given by

## \frac {r \sqrt{n-2}}{1-r^2} ##

EDIT: This is to test the hypothesis on whether there is no linear relationship as null , versus there is one, at your choice of confidence level.
 
Last edited:
Ranger Mike said:
I am seeking a very strong correlation coefficient and high confidence level.
Although correlation is often done for such measurements it is not a very good thing to do statistically. You should read the famous paper by Bland and Altman.
 
thank you fro the timely replies. Having spent all day revisiting the old calc book i am back up to speed on stats. Granted one correlation does not a decision but i think it will provide a lot of insight into our current metrology capabilities.
I have decided 61 sample size will be a good base line for 95% confidence per T table and my non scholastic customers like lots of samples to compare. I suspect the reading from all CMM and rotary instrument will be within a 1 micron range and i am betting on it being within 1/2 micron from past experience. Back in 1980 it was 2 1/2 micron!
would my data be of any benefit you the great people on this forum?
thank you
rm