Finding significance prior and after a process change

Question

Hi,

my customer has a chemical plant process which will be adjusted. At all they have 5 following process steps. To detect significance between the previous process and the process after a change they would like to deduct a certain process parameter. This seems to be Kind of common thing to do. I'm just wondering if this is something different than looking e.g. for the drift in two measurements.

Anyone with more insight is welcome

malcolm_moore1 · Answer

Hi Martin,

Looks to me like they want to test the effect of measurement equipment having adlusted for the effect of one or more covariates. If so, then one solution would be to use Fit Model to model the effects of measurement equipment and the covariates then look at the LSMeans differences table based on students t or alternative to test for differences in effects of equipment. The LSMean and error in these tests will be adjusted for the other factors.

Malcolm

martindemel · Answer

Hi all,

now I have more and new information.

1. In a running process, let's say a volume of material flows through different measurements instruments (in parallel) and you want to have the volume flow randomly but equally balanced through the different devices. Usually you would use an ANOVA to test if the measurements are really equally balanced or if there is one eye-catching device.

2. But in this process not everything runs randomly, e.g. one operater has a shorter distance to go than another. Or there are partly tools used at the beginning of the process, e.g. because of environmental circumstances. Those disturbance variables are well known. Now you want to know if e.g. one device has more measurements than another and maybe with higher values or is all fine balanced (excluding the disturbance Parameter you know). Does this come from the disturbance or is this something to act on.

3. E.g. For an assessment of the devices you would like to use the flow running through these tools, too, and still check if the devices run equally balanced. Therefore you would like to substract the effect of those disturbance from the measurement to have a fair comparison between the devices. An ANOVA would be biased by the disturbance effect ten.

The main question to answer is should I interrupt and adjust the process or is all fine. But all without additional testing. It's usually just one Parameter to substract and it is not a change in the process paraeters or the process itself, the disturbance is part of the process from the very beginning.

Looking on this from a different perspective you may would like to judge over the tools at the beginning of the process, but want to substract the variance of the differnt measurement devises.

Hope this makes it more clear.

Thanks,

Martin

Peter_Bartell · Answer

Martin: Is this some sort of transmitted variance type problem? That might be one way to approach the process characterization?