Test of Significance for Input Variables

[ontario]

Hi everyone,
I have a process that assembles components on an automated line. At the end of our line, we have a final test that measures variable "Y". We're having capability issues on "Y", so I've pulled the process database for the following input variables:
- Tester used (A or B),
- Pallet used (1 through 50),
- Part cavity number (1 through 4),
- A measured dimension (3.0mm - 5.0mm).

I have 100,000+ records in my database correlating the measured "Y" to each of these four inputs. My ultimate goal is to identify the specific cavity, pallet, tester (or combination) that is causing my outliers.

Any thoughts on where to begin?

 

[Bev D]

do you have JMP or minitab? either of these will be helpful in dealign so much data

I would start by plotting the y data (individual characteristic values) for each input factor. simple Y by X or multi-vari chart will do. (in JMP use the multivariate chart). It may be nice to add the means diamonds, but not necessarry; if you can't see a difference, then p values and confidence limits won't help you.

you might also plot the data in time series with the various factors nested within the time periods...

you could also post the data and we can show you how some of this is done...

 

[ontario]

Thanks for the input, Bev.

I have minitab. I'll start with X-Y and multivar plots to see what I can find.

Thanks again!