Newsgroup readers,
I am trying to clarify the resolution or resolving power of an analytical
method, calculated from the number of replicates and the za statistic.
Resolving power is not distribution. Distribution refers to the range of
expected test results, on the average, over the long term, following the
natural distribution of the test result population controlled by standard
deviation.
I believe resolution refers to the ability of a test method to detect a
statistical difference between two test results x1 and x2. It is the minimal
(statistically) detectable difference between 2 samples with similar analyte
concentration. It is a significance test.
Consider the test of 2 samples with similar analyte concentration, each sample
tested three times by the same test method, and the two averages compared. The
resolution of the test method (za = .05) under these conditions is 3.3 %. The
two samples are considered indistinguishable if they are less than 3.3 % apart;
and distinguishable if equal to or greater than 3.3 %.
Tabulating the resolving power of a method as shown below, allows the chemist
to compare samples and test results by simple inspection at any time. This is
easier for chemists who do not enjoy statistical hypothesis tests.
This is not to say that quality control chemists should not trend their data
and look for process and analytical method control problems, etc. Control
charts show atypical events quickly and clearly.
This question is posed to clarify the meaning and correct use of the resolving
power of a test method.
I would appreciate your comments to me at [EMAIL PROTECTED] and/or this
newsgroup. I am especially interested in literature references.
Thank you.
Stan Alekman
RESOLVING POWER AS A FUNCTION OF PRECISION AND SAMPLE SIZE FOR TWO GROUPS OF
EQUAL SIZE (za = .05)
s n = 1 n = 2 n = 3 n = 4 n = 5
0.1 0.28 0.20 0.16 0.14 0.13
0.5 1.4 1.0 0.80 0.7 0.6
1.0 2.8 2.0 1.6 1.4 1.3
2.0 5.6 4.0 3.3 2.8 2.5
=================================================================
Instructions for joining and leaving this list and remarks about
the problem of INAPPROPRIATE MESSAGES are available at
http://jse.stat.ncsu.edu/
=================================================================
