I'm the author of this piece. It describes a very simple experiment in which I counted how much change I picked up for a month, and my analysis of that data. It might be especially interesting if you're interesting in counting rare events or in cases where normal approximations aren't appropriate.
Interesting article, I admire your dedication to data collection! A zero-inflated Poisson model might be more suitable for your data, since there are a lot of days where you didn't find any coins. It assumes the data generating process has two stages: in the first there is a Bernoulli trial with probability of success p. In the case of no success then there are zero observed events. In the second stage, for cases where there was a success in the first stage, the number of events is determined according to a standard Poisson distribution. I think there's an R package to fit this kind of model.
