0
$\begingroup$

We use control charts in our lab to monitor process variation, and I'm trying to reconfigure the chart for one of our methods, so data is better represented.

For the method in question, we use a control standard and plot measured values in a control chart. However, the attribute measured in constantly decreasing (gas leaving solution). The change is slow, but constant.

Using control charts as we currently are (from the population), leads to calculation of a mean and control limits which aren't very representative of the process as it presently is.

I've tried recalculating the average using only past few data points, rather than the whole dataset, but I'm not sure whether the limits calculated from that mean are statistically valid, and the control limits are 'delayed' if that makes sense.

Changing the limits to include a subset of the dataset around a point (previous five and next 5, for example) gets rid of the delay, but we need to know immediately if a measurement is out of control.

Alternatively I've looked at using IMR charts, and watching the variation with those, but the individuals plot moves down and away from the mean.

$\endgroup$

1 Answer 1

0
$\begingroup$

I believe differencing is the solution I needed.

Using info from here: https://www.investopedia.com/articles/trading/07/stationary.asp

I was able to remove the drift from my process chart.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.