Looking for a Statistical Test

#1
My company has a process that produces a metallic alloy. One of the constituents of the alloy will vary from 0.10% to 2.00%+. However, the specification maximum is 1.00%. At the beginning of the year we were running a large percentage of the individual batches above the 1.00% limit and initiated a process change to correct the problem. We were not able to eliminate the out of spec batches, we will run periods with a low number of out of spec and other periods with a larger number.

Keeping in mind, we do not care how much above or below the 1.00% limit the actual analysis is (although I have that data), only that the batch is within or out of spec. Is there a statistical test that I can use to demonstrate that the process change has significantly improved the situation. I have more than a 1000 data points from 2006, and maybe 800 or so from 2007. All data is broken down by month.

If someone can just provide the name of the test and a possible reference, I should be able to figure it out.

Thanks in advance for your assistance.

emgall
 
#2
I guess the easiest thing to do would be a two sample test for the difference in proportions--ie before the change we saw x % percent out of spec and after the change we saw y% out of spec and the probability we'd see this large a difference given our data and the test assumptions is z. This is a very easy test that you can even do in excel.

Other than that there are a variety of tests based on linear models (anova, etc) that you can do. To do them correctly you have to really understand the data so you can set up the models correctly. You should also understand the hypothesis that you're testing and the test assumptions.

Dave