[Leica] Performance and statistical sampling

[Erwin Puts <imxputs@knoware.nl>]

Eric wrote:
>Yes, it's questionable, when they only test one sample out of the thousands
>of examples that were made of that particular lens. Such an impression by a
>test of that sort is, to say the least, questionable. It may actually test
>out way better than the average. My courses in statistics makes me way too
>cynical on this point I'm sure

Basically you are correct. If you would extend that argument to testing in
general: hardly a test ever produced in the worldwould be meaningfull. My
statistical courses tell me that a meaningfull sample should consist at
least of 20 items of the same product. Now it is impossible for me, and
anyone I presume, to test 20 boxes of the Apo-Telyt  3.4/135 to be able to
note significant deviations and find the margins of error in the mean
values produced.
All testing is done on the IMHO correct implication that quality control
now is sufficiently high to warrant general conclusions to be drawn from an
imperfect statistical population. A few weeks ago I had the opportunity to
test more than 25 lenses of Leica (old and new, all focal lengths, all ages
and versions)and I found not one lens to behave beyond the standards set by
Leitz/Leica.
So again this is only circomstantial evidence (Marc I need help here) but
it shows that testing beased on a very few samples (at least in the Leica
world) has some validity.

Dominique wrote:
>Pop Photo does not give the method
>they use. Moerover there is a "subjective quality factor".What's that ?

They did and quite extensively: in the 1990 issue they published an indepth
article explaining their method and (quite unique in the photographic
world) referred to a number of scientific articles to back up their method.
Whatever you think of their approach it  has scientific support.



Erwin