Can anyone point me to a useful reference for estimating a confidence
interval for the mean of a moderate (60+) number of observations for which
values 0, 2,3,4,5,6,7,8 are possible. I have several observed average
values for different conditions, and need to check whether they are
significantly different from one another or from a global mean.
Background:
A colleague has carried out an experiment in which she counted the
number of matching DNA profiles found after amplification using standard
heating blocks. After doing the experiment (*&%!*) she came to find out
what constituted a significant result.
She ran 8 profiles per run, spread (consistently) between three different
positions. The possible number of matches is 0, 2,3,4,5,6,7,8. (ie it doesn't
include 1). She now has average values for each position, and for the
average number of matches per run.
It may help to know that the match rate is quite high, so zero is a relatively
rare occurrence - c. 5%
BTW: there are only 3 distinct types of position on the blocks; a
chi-squared test on observed avergae for each against expected global
average is woefully insensitive with only 2df. I'm looking for something that
is sensitive to the number of runs, which is of the order of 60 runs, or 480
separate observations.
------------------------------------
Steve Ellison
Laboratory of the Government Chemist
Queens Road, Teddington, Middlesex
ENGLAND TW11 0LY
Phone: [ +44 | 0 ]181-943-7325
Fax: [ +44 | 0 ]181-943-2767
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|