I understand the difference, but in every place I've ever worked systems have already been set up to sample everything as an attribute using ASQ Z1.4. I have never actually used Z1.9 in practice.
I've given myself "homework" and done experiments with Z1.9 to understand it.

Is it just that attribute sampling is easier to understand? You reduce everything to either "in spec" or "not in spec"?
You don't have to deal with sigma or ranges?

In what cases would this not be advised?

In general it seems to me variable sampling requires less samples with a little calculation, where if you "attributize" those dimensions you need more samples but can use a go/no-go gauge.

please discuss.

I understand the difference, but in every place I've ever worked systems have already been set up to sample everything as an attribute using ASQ Z1.4. I have never actually used Z1.9 in practice.
I've given myself "homework" and done experiments with Z1.9 to understand it.

Is it just that attribute sampling is easier to understand? You reduce everything to either "in spec" or "not in spec"?
You don't have to deal with sigma or ranges?

In what cases would this not be advised?

In general it seems to me variable sampling requires less samples with a little calculation, where if you "attributize" those dimensions you need more samples but can use a go/no-go gauge.

please discuss.

Whether any feature in question is characterized as variable or attribute is a function of what you need to know. In some instances, you will only need to know whether a thing is black or white, big or small, fat or skinny. In other instances, you might need to understand the nature of the variation and control it more precisely. Deciding whether to use an attributes or variables sampling plan isn't a matter of (strictly) personal preference, or simple expediency. Decide what you need to know, then decide how to get the information you need.

Whether any feature in question is characterized as variable or attribute is a function of what you need to know. In some instances, you will only need to know whether a thing is black or white, big or small, fat or skinny. In other instances, you might need to understand the nature of the variation and control it more precisely. Deciding whether to use an attributes or variables sampling plan isn't a matter of (strictly) personal preference, or simple expediency. Decide what you need to know, then decide how to get the information you need.

Makes sense.

curiosity - have you used variable sampling very often in your career?

anyone - I'd appreciate an example from past experience where you did variable sampling rather than attribute.

Hi Discordian,

With variables sampling, generally smaller sample sizes are required, as compared to variables sampling.

For example, in ANSI/ASQ Z1.4-2003, for a lot size of 500, general inspection level II, code H, single sampling, normal inspection, AQL 1.0, the sample size is 50, with an accept of 1 and a reject of 2.

In ANSI/ASQ Z1.9-2003, for a lot size of 500, general inspection level II, code I, normal inspection, double specification limit, variability unknown, AQL 1.0, sample size is 25 with M = 2.86.

In Z1.9 pages 100 to 104 are comparison tables describing some of the differences. For my example, the 95%, 50% and 10% probabilities of accepting a lot have similar percent nonconforming values.

The disadvantage of the Z1.9 plans are they are more difficult to explain and use.

I had used a Z1.9 (MIL-STD-414) plan for acceptance of forgings when the supplier was having difficulty controlling a key dimension.

Wes R.