Sample size for creating a data base as a reference to a tested variable

[Jonathan]

Hello,

I am interested in creating a database of certain properties of a material in order to have a reference for checking products that got rejected by the user.
the intent is to have a range of values for each property in order to place the rejected products properties in that range.
My question is how do i determine a sample size to create a reliable database?

the purpose is to check for material degradation on the rejected products.
the range will be made up from products with new material.

by comparing the rejected materials properties to the range I prepared, I can have an estimate of the degradation that took place.

Thank you!

 

[Bev D]

We would need more information on. The situation to provide useful information: do you know the lot number of the returned material? Do you know the ‘range of material values at time of production or acceptance for shipment? What is the material? What form is it shipped in? How homogenous is the material property within a batch? From piece to piece?

Also if I’m understanding your description you want to take the returned material and ‘categorize’ it’s material property value into some ‘range’ instead of just recording the value? Why? If you know the lot’s original value (average and SD as long as the sample size used is representative of the non-homogeneity of the production batch/lot) and you know the returned value then you can easily calculate teh amount of degradation....

 

[Ronen E]

Hello,

I am interested in creating a database of certain properties of a material in order to have a reference for checking products that got rejected by the user.
the intent is to have a range of values for each property in order to place the rejected products properties in that range.
My question is how do i determine a sample size to create a reliable database?

Thank you,

the purpose is to check for material degradation on the rejected products.
the range will be made up from products with new material.

by comparing the rejected materials properties to the range I prepared, I can have an estimate of the degradation that took place.

Thanks again.

Are you planning to sample "products with new material" (new out of production?), in order to establish the property's range in new (acceptable) products?

 

[Jonathan]

Thank you for the response.

the material is PcU.

I have the LOT number but there is no information on the properties of the new produced PcU.
The LOT number enables me to know from what batch of raw material the product was made from.
post production the material is checked visually and therfore i dont have any information about how homogeneous are the material properties in a batch.

my assumption is that the visually approved products are accepted and now, I assume, the first step is knowing the sample size to take in order to get the missing information.


Thanks you,

 

[Bev D]

Why can’t you test the post production material? That is really the first step. The correct sample size is dependent on many things:
- the homogeneity of the original production/fabrication lot
- the standard deviation of the lot
- the amount of precision you need in your estimate
- the alpha risk (you might think of this as ‘confidence’ which equals 1-alpha and not is not confidence in the layman’s sense)

Generally the first step is to take a series of small samples across the lot to determine the first two items. You appear restricted to only measuring returns which makes understanding the first two things very difficult.

I would start by measuring all returns until I was comfortable with the data - when it looked ‘stable’.

There is another option: Why do you want this data? What are you going to do with this data? Are you trying to correlate one of hte properties with the reason for the return? In other words are you trying to do some kind of root cause analysis?

What do you think establishing a ‘range’ will do for you?

 

[Jonathan]

I am trying to characterize and evaluate the wear in order to do a root cause analysis.
since i dont have the original parameters of the post production materials i think establishing the range will help me in determining any degradation of the material and loss of mechanical properties.My idea was to start testing now for homogeneity and standard deviation with the products that will be out of production soon, just like you suggested.

i started to measure returns and to analyse data too.

I was interested to know how can i estimate the number of samples to check from the post production in order to establish those parameters.

 

[Ronen E]

@Jonathan, maybe start by reading these 2 articles in order:
Rational Subgrouping | Quality Digest
Rational Sampling | Quality Digest