[ | Next | Previous | Up ]

Re: Attribute Vs Variable Sampling

From: Stan Hilliard shilliard@samplingplans.com
Date: 15 Aug 2001
Time: 11:13:23



The kind data that you have will determine whether need to use an attribute sampling plan or variables sampling plan.

Attribute data is the simplest kind. You sample some number of items and you classify each item as either having some attribute, like being defective, or not. You will have some standard for what a defective item is -- like broken, scratched, non-functional, etc.

When your data points are measurements on a numerical scale you have variables data -- like weight, diameter, tensile strength, etc.

Variables data contains more information than attribute data per data-point. This is because it allows you to assess "how much" or "how bad" or "how good" rather than just "yes its defective" or "no its not defective".

Because variables data contains more information per data point than attribute data, variables sampling plans require fewer samples than attribute sampling plans -- for the same level of protection. This translates into lower material cost, less work, and less time for variables sampling plans.

It is particularly wasteful of information (and materials, and work, and time) when people go to the trouble of taking measurements (variables data) and then compare each data point to a specification limit and treat each point as either "in" or "out". (Attribute data).

This way of converting measurements to "in" or "out" downgrades information-rich variables data into information-sparse attribute data. A sampling plan that uses such go/no-go data is an attribute sampling plan.

The attribute plan decision rule will reject if too many points are "out". Typically, the maximum number of defectives allowed in the sample is calculated with the binomial distribution.

A sampling plan that uses the original measurements is a variables sampling plan. The variables plan decision rule will reject if the sample average of the measurements goes outside of some calculated acceptable range. Typically, the limits of that acceptable range are calculated with the normal distribution.

In general, most of the information that we receive (about anything) is in the form of either attribute data or variables data.

Stan Hilliard

Last changed: November 20, 2007