Search the Elsmar Cove!
**Search ALL of Elsmar.com** with DuckDuckGo including content not in the forum - Search results with No ads.

Determining of sample size for 'Operational Qualification'

#1
Hi, guys
I need your expertise for the issue I have with respect to getting the justification on the sample size of OQ
I plan on performing Tensile strength testing after some proccess. I have no idea how many samples are needed.
There are more than 3 parameters on the process,
If I need 3-5 samples for each experiment , it would be more than 200 samples depending on the combination of parameters.
I don't think the sample size for the OQ does matter. The IQ and OQ are really about the equipment. It's very inefficient and costly,

but I can't find any rationale for sample size of OQ
Any help for sample sizes for a OQ would be appreciated.
Thanks
 

yodon

Staff member
Super Moderator
#2
What I've seen in cases you describe is to use OQ to tune in the parameters that get you the best result (and establish the bounds of the parameters) and then do a PQ with those parameters (typically 3 runs). So sample size on OQ isn't critical in this approach but you still have to determine run size and sampling rationale on PQ.
 

Bev D

Heretical Statistician
Staff member
Super Moderator
#3
OQ is not intended to tune-in the parameters. it is intended to validate (or qualify, which is what the Q stands for in OQ) that the parameter settings (input specifications) will produce acceptable parts. The sample sizes may be small or large depending on the type of product and the study design. The sample size does matter. However, since you are typically running at the settings that will produce results at the max and the min of the output specifications (largest/smallest fastest/slowest, hardest/softest, etc) you can typically use continuous data statistics to determine the sample size, remembering that you are not trying to estimate an average or a standard deviation, you are trying to determine if the results are within spec. Of course if you are dealing with pure categorical data sample sizes will be higher.

It is typically not required to run every combination of the settings, only those that will produce the minimum and maximum results. Of course to do this you must understand the physics of the process...
 
#4
OQ is not intended to tune-in the parameters. it is intended to validate (or qualify, which is what the Q stands for in OQ) that the parameter settings (input specifications) will produce acceptable parts. The sample sizes may be small or large depending on the type of product and the study design. The sample size does matter. However, since you are typically running at the settings that will produce results at the max and the min of the output specifications (largest/smallest fastest/slowest, hardest/softest, etc) you can typically use continuous data statistics to determine the sample size, remembering that you are not trying to estimate an average or a standard deviation, you are trying to determine if the results are within spec. Of course if you are dealing with pure categorical data sample sizes will be higher.

It is typically not required to run every combination of the settings, only those that will produce the minimum and maximum results. Of course to do this you must understand the physics of the process...

Hi Bev D,
Thanks for your help
Where can I find detailed references or information about this

"It is typically not required to run every combination of the settings, only those that will produce the minimum and maximum results."

I really appreciate your help
 
Top Bottom