Originally Posted by
Cyclaholic
I can't comment on your dataset since you haven't published it but I don't think the binomial model is the correct one to use. I think you should model it as a Poisson distribution as the of a flat per unit mile (or 10 miles if you want to stick with that), then you can compare the lambda of each different tire and I think that would give you a much better indication of its flat resistance.
Well, statistics was never my strong suit but I don't believe the Poisson distribution works for me because I don't have enough data to know the actual probability of getting a flat for either tire with any kind of certainty at all, particularly for the tire that didn't get any flats. So what I'm actually doing is estimating the likelihood that the actual probability of getting a flat for the two tires is nearly equivalent. Of course, this doesn't tell me anything about how likely I am to get a flat in the next 10 miles with either tire, or even how many flats I would get on a second set of the same tires over 2000 miles. It just tells me that tire B
probably really is more flat resistant the tire A.