Sampling

When measuring a group of things, you may not be able to access every item or you may not want to measure everything. Sampling allows you to measure a few items and then generalize to draw conclusions about the whole population (all possible items that are of interest).

In social research, you cannot interview everyone, so it is common to take a sample of the population. In quality assurance, such as in acceptance sampling of inputs (or final test checking of outputs) the cost and time constraints make sampling a reasonable option rather than checking everything.

Sampling by attributes uses counting of conforming / non-conforming items. Sampling by variables is used in continuous measurement.

Sampling may also be necessary if testing destroys the items being sampled, for example where strength testing applies mounting pressure until the item snaps.

The standard question in sampling is 'How many do I have to sample?' Unfortunately there is not always an easy question to this. Sample size generally depends on the confidence you seek and the likelihood of finding a problem. One of the simplest ways is to use sample size tables which are created to suit the particular sampling context.

Sampling

See also: