Originally posted by mtkeller My argument is that the conventional wisdom around here about why these solenoids fail is based on nothing but rumor and innuendo. People keep making statements (such as these things never fail in the first x years or that they only fail after prolonged periods of disuse or that they don't fail after y years) that are not backed up with anything. If one were to undertake a study, they would formulate a null hypothesis and look for evidence to disprove it. The null hypothesis would be that there's some distribution of times before failure for these solenoids (likely, one would hypothesize a particular distribution and parameters for the distribution). You'd then collect a random sample of cameras using the solenoid and see at what time the solenoid failed (or if it is still operating without having failed). No one has done that sort of study.
This 'discussion' is almost like having
photogem play International-style "football" while you play American-style "football".
He is deducing general behavior from behavior of individual cameras - going from individual behavior to group behavior - while you use want to use statistics to get group behavior.
This is really not a discussion of "rumor and innuendo" - individuals are trying to learn from anecdotes, because - honestly - that is all we have.
I have taken statistics at four different Universities - around a dozen courses in all - but no one has ever talked about selecting the 'correct' statistical model, which is more art than science.
Statisticians often use the 'Normal model' for reasons of theory and practicality, but we already have evidence that it would be the wrong selection in this case.
Perhaps the 'Exponential model' would work better with its non-existant tail towards zero and its long tail towards infinity {it's usually used for light bulbs afterall},
but collecting a random study would be hopeless in any case - we have already had many discussions about the bias that comes from owner's reports.