Performance Characterization of EDA and Its Potential to Improve Decision Making in Product Batch Release - Good Cascade Impactor Practices, AIM and EDA for Orally Inhaled Products - page 218

Table 8.9 Simulation results for various scenarios showing type I and type II error rates for the
two approaches and the relative type I error rates for grouped stages compared to EDA; results
from groups 2-4 are combined to yield the overall grouped-stage decision (i.e., if any of the groups
2-4 fails, the overall decision fails)

Error rates Relative type I

error rates:
grouped stages
versus EDA

Grouping
distribution
percentiles

Prediction
interval
(%)

LPM / SPM ratio Grouped stages

Type I Type II Type I Type II

10, 90 95 6.57 0.03 22.82 0 3.47

99 10.86 0.01 30.51 0 2.81

99.9 17.01 0.01 41.82 0 2.46

5, 95 95 8.48 0.04 23.08 0 2.72

99 13.89 0.01 32.39 0 2.33

99.9 19.66 0.01 43.09 0 2.19

1, 99 90 9.61 0.09 20.95 0.09 2.18

95 12.24 0.08 26.05 0.02 2.13

99 17.38 0.05 39.03 0 2.25

99.9 20.81 0.02 48.04 0 2.31

Fig. 8.49 OIP product w9k001 : OC curve ( solid green line ) and results for LPM / SPM ratio; black
symbols are correct decisions, blue symbols are false rejections (type I errors), red symbols are
false acceptance (type II errors). LAL and UAL are lower and upper acceptance limits,
respectively

In addition to the relative rates of false rejections and false acceptances, the
pattern of incorrect decisions is also important in assessing the decision-making
capability of a method. If the errors are near acceptance limits, but a method is
clearly discriminating in the majority of acceptable and unacceptable regions, that
method could be considered more useful in making correct decisions than the one