Chemistry Reference
In-Depth Information
tAble 6.2
summary of statistics for testing of outliers
Outlier
: A value in a set of observations that is so different from the rest that it is considered a member
of another set or population.
The most common method for testing of outliers is probably Dixon's Q test
a
, which reads as follows:
Q = [suspect value − nearest value]/(largest value − smallest value)
where [suspect value − nearest value] represents an absolute difference, no sign is intended (±). The
suspect value is always that number which is farthest from the mean or average value of the set.
Critical values for Q for any probability value (
P
= 0.05 and
P
= 0.01) are found in various tables.
a
For example: Th.e following values were obtained for the nitrite concentration (mg/L) in a sample of
river water: 0.403, 0.410, 0.401, 0.380
The last measurement is suspect; should it be entirely rejected?
We have
Q = [0.380 − 0.401]/(0.410 − 0.380) = 0.021/0.030 = 0.7
From any table of Q values, the critical value for Q is 0.831 (
P
= 0.05). Because the calculated value of
Q does not exceed this, the suspect measurement should be retained.
a
J.C. Miller and J.N. Miller,
Statistics for Analytical Chemistry
, Halsted Press, Division of John Wiley
& Sons, New York, 1986.
6.15 conclusIon
Many of the concepts and tests in this chapter are important in the development of
harmonized specifications. They are not universally applicable, or necessarily all-
encompassing. Tests other than those listed here and in the guidelines may be needed
in particular situations or as new information becomes available. New analytical
technologies are constantly being developed, and their use is always encouraged
where justified. In general, proposals to implement the concepts outlined here and in
more detail in the guideline should be justified by the applicant and approved by the
regulatory agency prior to implementation.
Whether the topic is setting acceptance criteria, out-of-specification results, sys-
tem qualification, or method validation, the discussion eventually turns to suitability
and acceptability for intended use. And as with all validation topics, the common
denominator is doing good science.
reFerences
1. ICH Q6A: Specifications for new drug substances and products: Chemical sub-
stances.
International Conference on Harmonization of Technical Requirements for
the Registration of Drugs for Human Use
, Geneva, Switzerland, March 1995. See also
www.ICH.org under topic of Quality guidelines.
