Sample Size for the Diagnosis of Conservation Units

Abstract: Use of the phylogenetic species concept in defining conservation units is based on the assumption that the fixation of a particular character state in a population is diagnostic of a long history of reproductive isolation. In practice, diagnosis is usually based on the character states of a small sample of individuals rather than the states of the entire population. Unfortunately, when sample sizes are small, samples in which all individuals share one character state can easily be drawn from populations that are actually polymorphic. I describe statistical methods for examining how much confidence can be placed in the diagnosis of a conservation unit, given the operative sample size. The methods estimate the probability of drawing a sample in which all individuals show the same state, if individuals with unsampled ( hidden) states actually exist in the population at some hypothetical frequency (e.g., 0.05). I considered finite and infinite population-size models. The infinite population-size model suggests that in order to reject with 95% confidence the hypothesis that 5% of individuals carry hidden character states, a sample of 59 individuals is necessary. Finite population-size models give slightly smaller critical sample sizes for diagnosis with 95% confidence. I describe methods for including the effect of uncertainty in estimating population size when calculating critical sample size, and I discuss extensions to multiple characters and the impact of spatial structuring of character states. My results suggest that confident diagnosis requires sample sizes much larger than those commonly used when the phylogenetic species concept is applied to defining conservation units.