COS 92-8
A test for a species range shift based on stochastic dominance
Climate change is expected to lead to shifts in species distributions. Data on these shifts have been collected and analyzed across many taxa. However, existing tests for distributional shifts may be sensitive to distributional changes that do not constitute an unambiguous shift. We propose stochastic dominance as a non-parametric, unambiguous indicator for shifts of population abundance distributions. We developed a double-bootstrap method which reliably detects dominance between distributions. We explain our method, illustrate its performance with simulations, and illustrate its usage with cod distributional data over the past forty years.
Results/Conclusions
We present results from a simulation experiment aimed at assessing the performance of the test. These results indicate good agreement between the nominal and actual significance level when the null hypothesis is true and high power when the alternative hypothesis is true provided the magnitude of the shift and sample sizes are not too small. We also present an application of the test to determine whether the distribution of Atlantic cod (Gadus morhua) in the Northwest Atlantic experienced a northward shift between 1985 and 1995. The test found no evidence of such a shift (p = 0.009).