Previous investigations of geomagnetic reversal sequences have shown that a nonstationary gamma distribution provides a good description of the intervals between reversals. The general problems of parameter estimation and testing for differences between the statistical properties of reverse and normal polarity sequences are considered. A likelihood method is suggested to overcome the problems of nonstationary and the consequent dependence of polarity interval lengths while avoiding the problems associated with the use of sliding windows. Previously, the good fit of a gamma distribution has been interpreted as implying that geomagnetic reversals are a gamma process. An alternative possibility is presented here by showing that incomplete data from a Poisson process leads to a distribution of interval lengths essentially indistinguishable from a gamma distribution.