Originally posted by JMS I disagree. You need to work on your statistics.
Let the true population distribution of the AF speed at a given light level with a given AF unit and voltage supplied have mean 1 and standard deviation 0.05.
Then, let the true population be estimated by a study with sample size n=100 tests.
The distribution of x-bar will have mean 1 and standard deviation of 0.005. Thus the probability can be easily calculated that if PopPhoto's tests are accurate, there would be approximately a 95% chance that a second test will yield the same results as the first at each light level if exact to the .01 level. At 9 light levels total, there is approximately a 63% chance of identical results.
Hello JMS!
I definately have to work on my statistics and as well as on my English as both aren't my special subjects. So I'm not sure if I could follow you on the first part of your posting. Do you suggest that Popphoto repeated the focussing test on each light level n=100 times and then averaged the numbers? If true I will agree that with equal hardware and equal circumstances such identical results might be possible. However I assumed a single testing on each light level.
I've just revisited the
results and found an indication that makes my point quite clear: Whereas the focusing speed gets slower for every lower light level, the number for EV8 is in fact lower than the number for EV10. So here the K10D focuses faster, if the ambient light falls by 2 stops?! This is perfectly possible with a single measurement and nothing to worry about. But of course this error should disapear with higher n. And how about the chance that this anomaly recurs at the same step (EV10 --> EV8) with the K20D on a controlled test 13 months after the first one?
The main point of my first post was the practical improbability of the results, assuming one testing on each light level. N must be very high to get equal results with an equal setting. And this setting is obviously not equal. I hope you can agree on this.
P.S.: I'm sorry for any misunderstanding as a result of my modest English and my rather intuitive approach to statistics.
Have a nice day!