From: "Ulhas Deshpande" 
To: "ATM" 
Reply-To: "Ulhas Deshpande" 


Hello All
Mark Holm in response to John wrote
>>You are right to be concerned about the variability in your data.
Practice
can
improve it somewhat, but you will probably find that, when you are working at
the level needed to confidently get a Strehl ratio of 0.9 or higher, the
measurement uncertainty will continue to be larger than you would like.  In
this case, the only recourse is more testing and use of statistics (chiefly
average or median) to pull truth out of the data.  Testing in more zones at
least partly compensates for noisy data too.
>>
I also seem to be suffering from this malaise. On first test I got Strehl =
0.918 ( I posted the SixTests file to the group couple of days ago). Just
to make sure I did another set of four readings per zone and this time I
got S= 0.877 ? I plan one more today.
>>
 One of the beauties of making multiple readings in each zone is that
probability predicts that readings serioulsy in error should be rare.  If you
make several readings, the average should quickly approach the true value. For
small sample sets, statisticians say that the median is even less likely to be
affected by noise than the average (mean).  The sample size has to be at least
3 for there to be any difference between the mean and median.  (That is 3 per
zone.)
>>
Will it make any statistical sense if readings from tests carried out at
differant times and perhaps with slight changes in tester distance are
combined and averaged or medianed(?)

Trying to get more consistent results Ulhas

--- BBBS/NT v4.01 Flag-5