I will try to explain to the best of my abilities.

My coworker and I has been tasked with finding out how a turning machine in our production facility performs, but we have do this retrospectively - if possible, we will use data from very recent productions.

The machine is running for days - and during that time two items will be taken and meausured in succession every few hours. We have been discussing removing operator influence, along with other external influences. My coworker got the idea that since the two samples are taken right after each other, we could assume that they should be alike, and that any difference would be due to influences. I guess this sort of makes sense.

*Each point is comprised of the mean of two measurements, one for each item (the two items are produced right after each other).*

So, if this is a viable approach, how would we go about it? Let's say that we have 100 samples, where each sample is comprised of n = 2. Our statistics system logs both measurements on a point on a chart, with the mean of the two measurements on the line between the values (see picture above). This then repeats for the next samples every couple of hours. I think it may be difficult to obtain removal by all influences retrospectively, but how would we "deduct" the influence and how do we know what the influence is?

Thanks for your time,

Nicholas