Biomedical Engineering Reference
In-Depth Information
Fig. 8.38
Definitions for type I and II error regions in relation to data presented in Tables
8.7
and
8.8
8.5.2.4
Results from “Christopher-Dey” Strategy
Christopher-Dey compared the EDA and grouped-stage approaches statistically using
multivariate modeling of the APSD profiles. Assessment of the relative capability of
the two methods to correctly detect shifts in the APSD via shifts in
MMAD
was based
on misclassification due to false rejection rates (type I errors) and false acceptance rates
(type II errors) using simulations of multivariate non-normal data and OCCs.
Shifts of mass within a distribution or shifts of the entire APSD will necessarily
change its
MMAD
. The ability of a given approach (EDA or grouped stages) to
detect shifts in APSD can therefore be assessed by their ability to correctly deter-
mine whether
MMAD
falls within an acceptable range of
MMAD
values. The same
range of acceptable
MMAD
values was used to evaluate each metric in order to
ensure an equitable comparison of the two approaches.
The two sets of metrics as well as the corresponding
MMAD
value (see Fig.
8.39
)
were each calculated for the CI-measured APSD data from each OIP (i.e., the iden-
tical CI data values were used to calculate each metric). In the EDA approach, the
focus was on the magnitude of the
LPM
/
SPM
ratio. In the grouped-stage approach,
the focus was on the mass of API assigned to stage groupings 2, 3, and 4 (Table
8.1
).
The particular EDA and grouped-stage metrics are illustrated side by side in
Fig.
8.40
, together with appropriate size boundaries for the ACI, when operated at
28.3 L/min.
For an outcome indicating that
MMAD
is within the acceptable range of values,
each of the grouped-stage results must fall within established acceptance limits. The
ability of the grouped-stage approach to make the correct decision as to whether
MMAD
is within acceptable limits is directly related to how well the combined
Search WWH ::
Custom Search