Quote Originally Posted by Gerald C Koch View Post

In order for the graphs to be really meaningful and to draw conclusions there nust be at least three seperate measurements for each data point. This means 3 trials for each developer noted on a graph. Then the various points can be shown with error bars. It is only after this that one can be confident of the relative placement among the curves. It is possible that the curves for Xtol and your developer to fall within each others confidence limits.

The following site may be helpful http://www.pafaculty.net/biology/kei...Deviation.html. It's a lot of work but necessary.

That's the advantage a corporation has: They can have engineers and testers work full-time on a project. Those of us doing this as a hobby in evenings/weekends take short-cuts. My short-cut has been to develop just once, and if all looks normal, accept that and proceed to the next test. Also, I have tight control over temperature, exposure, agitation, etc. to try to improve the chance that my first try will not have any anomalies. But your point is that there is still test-to-test variation, and only multiple tests will reveal the variation-ranges.

Do you have any ideas of how to get variation-ranges (error-bands) without tripling the amount of effort? Maybe if I make three runs of the most popular films, and just one of the others? Or three runs of anything that looks odd? Or maybe others could volunteer to share the tedious testing effort?