Advances and Applications in Statistics
Volume 2, Issue 1, Pages 19 - 30
(April 2002)
|
|
EVALUATING THE BOOTSTRAP p CONTROL CHART
Jann-Pygn Chen (Taiwan)
|
Abstract: The
techniques used in Shewhart control charts assume
that the observations are independent, and that the
statistics of interest are approximately normally
distributed. However, these assumptions are often
violated for the attribute data. In such cases the
control limits may not be accurate, and bootstrap
methods, requiring no a priori parametric
assumptions, may be useful. In this study, the
standard bootstrap (SB), the percentile
bootstrap (PB), and the biased-corrected
percentile bootstrap (BCPB), used to
construct control limits for fraction nonconforming,
were examined for different sample sizes. Their
performance and that of Shewhart control limits were
compared by simulation. The results indicate that
the conventional Shewhart control chart should not
be used to monitor the process quality
characteristics, and the PB method performs
better than the other two bootstrap methods.
|
Keywords and phrases: bootstrap method, p chart, ARL, Type I error. |
|
Number of Downloads: 14 | Number of Views: 1550 |
|