Random sampling
Technique that insures each item in a sample
for inspection is selected
completely by chance to be measured.
You should never do 100% inspection in any six
sigma quality plan.
Sampling is more accurate in statistical
process control for several reasons including it is less susceptible
for 'inspection fatigue' in a process, because
there is a lesser chance for errors. Products that are 100%
inspected sometimes are found to have a astounding number of defects.
By using SPC and other control
charts with a control plan, you can
measure fewer production piece parts, (typically using a systematic
sampling plan in unison with stratified
sampling insuring that sampling
bias does not become an
issue), thus saving time and labor, and in theory, still have the same
confidence level of the quality of you process.
Famous errors and lessons learned in random
sampling can be attributed to Literary Digest magazine in 1936. They took
a random sample from the telephone listings for a pre-election poll. The
10 million samples taken during the depression years were not accurate
because in 1936 the people who could afford a phone and magazine
subscription did not constitute a random sample. The company was soon out
of business.
Also see systematic
sampling, stratified sampling,
and sampling bias.
