Determining Sample Size for FDA Verification and Validation Activities

Q

QualityNewbie

Hello,

I'm trying to figure out how to determine sample size for our upcoming V&V activities. Is there a sampling plan that is widely used? I've heard LTPD is common, but was wondering if there were other methods in use.

Thanks in advance for your help!
 
C

Craig H.

Re: Determining Sample Size

Hello,

I'm trying to figure out how to determine sample size for our upcoming V&V activities. Is there a sampling plan that is widely used? I've heard LTPD is common, but was wondering if there were other methods in use.

Thanks in advance for your help!

Could you define what "V&V" is, please?
 
Q

Qualqueen

Re: Determining Sample Size

Could you define what "V&V" is, please?
Verification and Validation activities:

· Data verification is conducted to ensure that the data selected is the most appropriate for the application and is properly prepared for the product.

· Data validation is conducted to ensure that the data accurately represents how it will be simulated.

I think this is correct
 
Last edited by a moderator:

Ajit Basrur

Leader
Admin
Re: Determining Sample Size

Hello,

I'm trying to figure out how to determine sample size for our upcoming V&V activities. Is there a sampling plan that is widely used? I've heard LTPD is common, but was wondering if there were other methods in use.

Thanks in advance for your help!

Welcome to the Cove :bigwave:

Can you pl give more details on what are you validating ?
 

Bev D

Heretical Statistician
Leader
Super Moderator
Verification and validation testing is NOT the same as acceptance sampling (AQL, LTPD, etc.)

I have attached a very brief word doc with sample size formulas that can be used for V&V testing. (It is not all inclusive and does not have the utilization of matched pair testing which can be very powerful for rare events or small differences with reduced sample sizes.


If you post some of your examples we can provide much more specific help on this under discussed topic.
 
A

achorste

Thank you very much for helping on this issue Bev.

Maybe I haven't quite got the idea for determining 'delta' - the amount of accuracy that you want - does anyone have a simple worked example they could share? Primarily I'm looking at continuous data.

Many thanks for any help you could give,
 

Bev D

Heretical Statistician
Leader
Super Moderator
The "delta" is simply the amount of accuracy you want. You can think of it loosely as the size of a confidence interval. If you are going to compare two things experimentally and you would expect to see non-overlapping confidence intervals if there was a real difference then you can take teh difference between the two as 2Delta...

here is an example of how to use 'delta' in validation. We had a plastic device that had a baseline breakforce average of 19.7 and a standard deviation of 1.81. We wanted to validate two new platic formulations and test their break force to the current process. The specification was 15-25. There was almost no lot to lot variation so we could ignore any lot to lot complications in our sample size determination. We wanted to be able to detect a 1 pound change in force (the spread of the process took almost the entire tolerance range so even a 1 pound change would increase our proportion out of spec.) We selected the standard values for alpha (5%) and beta (10%). We included Beta because we wanted to be sure that we saw a difference as small as 1 pound.
The formula: n = [{(z_a/2 + z_b/2).SD}/Delta]^2 = [{(1.96 + 1.645)1.81}/1]^2 = 43.

We tested three lots; the current plastic and the two new plastics, each test run had n=43. There was no statistically significant difference bewteen the three lots.
 
A

achorste

Perfect - thank you so much Bev.

Determining a statistical rationale for sample size was an issue I've been struggling with for a while now - I think I've got it sorted!
 
L

Lorenzo Lopez

Dear Bev D,
Thanks for your help because my question (see below) is similar to the one from this thread's originator.
Regards,

Dear friends at Elsmar:
I'm the QE for R&D for a Type I product, helping my Marketing partner to justify the number of user representatives participating in a validation run for a new version of the user manual of our product (our product is software-driven and manual is about 200 pages long); hints, article links/attachments and direct comment from your valuable experience is highly welcomed !
Regards,
Lorenzo Lopez
 
Top Bottom