Biomedical Engineering Reference
In-Depth Information
The statistical approach presented in the MSA section uses regression techniques
to relate the
LPM
/
SPM
ratio and stage groupings to
MMAD
. This, in turn, allowed
an estimation of measurement system variability of either
LPM
/
SPM
ratio or
grouped stages and a comparison of this variability relative to changes in
MMAD
.
8.4.3
OCC Approach: Definitions and Basic Concepts
The concept of operating characteristic curves derives from the view of testing as a
decision-making process. Typical pharmaceutical quality control (QC) testing
involves making measurements on representative samples of a batch and using the
results as the basis to judge the quality of the material being tested either against
some predetermined quality standard (which is conceptually and statistically the
preferred way) or against a predetermined set of rules (which has been regulatory
practice to date) [
9
]. This assessment of the quality is used to decide on some action,
i.e., release or reject the batch. In the statistical sense, such decision-making is a
form of hypothesis testing. As with all hypothesis testing, one strives to design the
statistical test to be as reliable as possible in terms of mitigating decision-making
error, within the practical constraint of obtaining a sufficient number of sample
measurements. In the presence of some level of measurement noise and variability,
however, a finite probability of making an incorrect decision exists. The terminol-
ogy for corresponding error types is illustrated in Fig.
8.4
.
Type I and II errors (which could also be considered “misclassifications” rather
than errors) arise out of the uncertainty in the estimation of the true value of a qual-
ity attribute associated with the batch being tested. In general, there are two ways to
Fig. 8.4
The product QC decision process, possible outcomes, and potential error types
Search WWH ::
Custom Search