Is there a standard for sample size during R&D phase

Stefanos Papadakis

Involved In Discussions
Hello eveyrone,

I hope you are all doing well.

I would like to know if there exists a standard/paper/method for calcualing sample size during R&D phase.
I work at product testing and I would like to know how many samples during pilot phase need to be tested.

I am aware of the standards regarding production sample size but it does not work for me.

Thank you in advance
Stefanos
 

Jim Wynne

Leader
Admin
What standard are you working under (ISO 9001, e.g.)? What type of device? What has been your company's practice in the past? There are no standards, as such, that I know of.
 

Stefanos Papadakis

Involved In Discussions
We are working under ISO 9001 and as product testing dpt. we are part of the R&D team. When we test pilot parts we might have from 30 to 300 samples. So what I am asking is, if there exists any good practices or any ASTM, ISO etc. standards referring to sampling.

The tests we are condacting are dimensional tests and properties of the part under various stresses.
 

Jim Wynne

Leader
Admin
We are working under ISO 9001 and as product testing dpt. we are part of the R&D team. When we test pilot parts we might have from 30 to 300 samples. So what I am asking is, if there exists any good practices or any ASTM, ISO etc. standards referring to sampling.

The tests we are condacting are dimensional tests and properties of the part under various stresses.
You can get an overview of statistical sampling plans here: What kinds of Lot Acceptance Sampling Plans (LASPs) are there? For more details and sampling tables, go here: ASQ/ANSI QUALITY STANDARDS Z1.4 & Z1.9. There is an old military standard, MIL-STD-105E, that I've attached here. While it's been withdrawn, the information and tables are still valid and you might find it useful.

As you might gather, acceptance sampling is a fairly deep subject. It's been discussed a lot here, and you can browse many of the threads here. Feel free to post back with questions.
 

Attachments

  • MIL-STD-105E.PDF
    5.5 MB · Views: 243

mmasiddiqui

Involved In Discussions
Hello eveyrone,

I hope you are all doing well.

I would like to know if there exists a standard/paper/method for calcualing sample size during R&D phase.
I work at product testing and I would like to know how many samples during pilot phase need to be tested.

I am aware of the standards regarding production sample size but it does not work for me.

Thank you in advance
Stefanos
What is the AQL (acceptable quality limit) for your customer?
 

Stefanos Papadakis

Involved In Discussions
You can get an overview of statistical sampling plans here: What kinds of Lot Acceptance Sampling Plans (LASPs) are there? For more details and sampling tables, go here: ASQ/ANSI QUALITY STANDARDS Z1.4 & Z1.9. There is an old military standard, MIL-STD-105E, that I've attached here. While it's been withdrawn, the information and tables are still valid and you might find it useful.

As you might gather, acceptance sampling is a fairly deep subject. It's been discussed a lot here, and you can browse many of the threads here. Feel free to post back with questions.

Thank you for the reply, I will check the threads and the standards you proposed.
Sadly MIL-STD-105 does not work for us.
 

Bev D

Heretical Statistician
Leader
Super Moderator
First – there are no simple ‘look it up in a table’ answer in the quality or statistical sciences. These are complex topics that require deep understanding and thought…



In design V&V and pilot lot testing acceptance sampling sample plans are not appropriate (most of the time). The reasons for this are:

  • Most of the time people think about the categorical data sample plans which are based on the number of defects or failures. Yet in development we are more interested and much better served to measure continuous data against requirements to ensure that our design has enough margin.
  • Acceptance sampling plans only tell you that you haven’t exceeded some max limit but they don’t tell you how close you are, which again in design the real point.
  • Since most designs require fairly low to non-existent failure or defect rates the sample sizes get really large.
So we are much better off using continuous data and the formulas for estimating averages and standard deviations rather than the acceptance sampling tables. These formulas were intended to estimate actual performance which is the intent of V&V and pilot testing.

Directed testing and testing at the extremes are much more effective and allow much lower sample sizes that are not comprehended at all in acceptance sampling tables…
 

mmasiddiqui

Involved In Discussions
Thank you for your answer but at this phase (pilot phase) we do not have AQL nor customers. Lets say that we are our customers. :)
I agree with you, but I would still like to establish a AQL internally for pilot phase. The other way is to take the lowest AQL of the potential customer who will be using your product.
 
Top Bottom