Robust Test for Two Data Sets - Tooling on an Injection Molding Machine

R

rmacken

Hi guys,

I am currently developing a process improvement with regard to tooling on an injection molding machine.

I am proposing to run a set sample size before and after the change and compare the two results for defects on the molded part. However the defect rates could be 1.5% before and 1% after. I am trying to identify a robust enough test for significant difference.

The data is also attribute data and is not normally distributed. Is the Mann Whitley Test sufficient or do I need to be comparing proportions? If I do this will drive my sample size to an unworkable level.

Any help would be greatly appreciated.

Regards,
R.
 
Howard Atkins

Howard Atkins

Forum Administrator
Leader
Admin
rmacken said:
Hi guys,

I am currently developing a process improvement with regard to tooling on an injection molding machine.

I am proposing to run a set sample size before and after the change and compare the two results for defects on the molded part. However the defect rates could be 1.5% before and 1% after. I am trying to identify a robust enough test for significant difference.

The data is also attribute data and is not normally distributed. Is the Mann Whitley Test sufficient or do I need to be comparing proportions? If I do this will drive my sample size to an unworkable level.

Any help would be greatly appreciated.

Regards,
R.
Click to expand...

Welcome

What sort of defects are you finding?

What are your current tests?
 
R

rmacken

Our defects are with regard to the resolution of a material, so its either pass or fail as per our specifications. Currently we have an average resolution failure rate of 1.5% of each lot and we are looking to reduced this by making some tooling chages at the injection molding stage. Since this defect rate is so small I am trying to find a test that can show a slight decrease or increase in the defect rate. We are currently not using any set statistical tests as we have not got the confidence in them. We are just monitoring our results at the resolution test stage.

Thanks for the help.
 
You must log in or register to reply here.
Top Bottom