Stat question, and sampling
Six Sigma – iSixSigma › Forums › Old Forums › General › Stat question, and sampling
- This topic has 6 replies, 5 voices, and was last updated 13 years, 11 months ago by
WayneJ.
-
AuthorPosts
-
June 16, 2008 at 4:42 pm #50300
Im hoping someone smarter than me in stats can help me with this.Our company has a device that will measure 100% of the product real time. The measure of interest is a maximum requirement. The 100% measuring equipment can record the individual max data points for the lot. Our customer does not have this equipment but does have a similar method of measurement, and samples 7 units from each lot. They have agreed that our company can use our 100% inspection device to accept product, but they will also perform receiving inspection for the time being. What worries me is the high sampling error associated with such a small sample. Our customer will sample 7 units, estimate a mean and standard deviation and look at the 3 sigma distribution.How can we determine acceptability based on the customers sample distribution?Can we compare the two distributions and determine if the sample mean could be from the population? We are both looking for a solution for this.
0June 16, 2008 at 4:45 pm #172864Only review your customers Pp and Ppk data
0June 16, 2008 at 4:59 pm #172869Ron,Thanks for the response, but what is your reasoning?
0June 16, 2008 at 7:26 pm #172875Hi Waynej,
why don’t you just calculate the confidence intervals for the measurements of your client?Your value should be inside their interval in about 95% of the cases. If this is true the two measurements are consistent. Of course the confidence interval will probably be quite large with only 7 data points :)Regards
Sandor0June 16, 2008 at 10:28 pm #172882Plot the customers data on a X-bar & R chart. You need 25 samples to start the chart. The average of the chart should be very close to your data. The control limits should be close to your +/- 3 sigma.
0June 18, 2008 at 12:28 pm #172942Waynej, what is the rational behind 7 samples that they pick up from each lot. You have all the reason to be worried. If your customer understands statistical concepts then you will have to face less challenge. Your curve and customer’s curve would be very close to each other in a long run. Even for it to be so, right sampling at 95% significance level would be a bare minimum.
0June 18, 2008 at 11:48 pm #172958No idea why they picked 7 as a sample size. But, I had a meeting with them today and they agreed to accept our 100% data and statistically compare their small sample, until they gain more confidence in our method.It was a good day.
0 -
AuthorPosts
The forum ‘General’ is closed to new topics and replies.