the simplest model extending the assumption of constant
population size. Two approaches are evident. The first is by an
extension of the Wright–Fisher model to the case of variable
population size based on transformation of the time variable to
reflect the local rate of change of population size (Griffiths and
Tavaré, 1994). This approach retains the assumption that overall
population size is deterministic, and that the Wright–Fisher model
describes the relationship of one generation to its successor. This
approach is conducive to efficient simulation and to a considerable
degree of analytic study. However, it potentially leads to inaccurate
estimation of population parameters due to underestimation of
the variability in population size, particularly in the early stages of
population growth (Stadler et al., 2015). An alternative approach
(O’Connell, 1995; Cyran and Kimmel, 2010) is to use the theory of
branching processes, and in particular Galton–Watson processes,
to model populations evolving according to a common probability
distribution for the number of offspring on the basis that this
distribution is the same for the entire population and in each
generation.
It is of interest to compare the effects of these population
models (constant-size Wright–Fisher; deterministic growth; and
fully stochastic growth) across various population types. As we
shall see the key population parameter to be considered depends
on both population size and growth rate.
The structure of this paper is as follows. The stochastic model
used throughout, which is equivalent to that of O’Connell (1995),
is set out in Section 2. Section 3 contains a comparison of the model
with Wright–Fisher for the case of zero mean population growth,
with emphasis on fixation probabilities and times. The case of
supercritical growth is analysed in Section 4. Implications for the
loss or otherwise of heterozygosity under different parameter
regimes are considered in Section 5. An application to the
estimation of the time elapsed since the life of mitochondrial Eve
and human population size during her lifetime is given in Section 6.
Conclusions are drawn in Section 7, and an Appendix is devoted to
critiquing O’Connell’s analysis of the mitochondrial Eve problem.