APPLYING NEUTRAL THEORY TO THE FOSSIL RECORD: GUIDELINES FOR ESTIMATING HUBBELL'S THETA
We simulated abundance distributions to test when estimates of theta are unbiased by sample size or effort. To do this, individuals from an existing data set were randomly selected to generate communities of increasing size. We estimated theta for each randomly generated data set using the untb package in R and Etienne’s 2007 and 2009 likelihood methods.
Both of Etienne’s methods are unbiased by sample size and agree well. Given this similarity and the long run-times for large data sets for the 2009 method, the 2007 method is preferred when estimates of Hubbell’s migration parameter m are not necessary. Estimates of theta in R’s untb package vary with sample size and are unreliable.
For statistically robust estimations of theta, sampling should be designed to include many small samples, as opposed to a few large samples, as theta for a single sample underestimates the theta for a multi-sample data set. More samples provide repeated sampling of a metacommunity and improve the recovery of rare taxa. Consequently, variability in theta increases with smaller total sizes of a data set. Larger data sets often improve sampling of the metacommunity and lessen the effect of patchiness on observed abundance distributions. Small data sets collected from communities that have a high degree of patchiness are unlikely to represent the metacommunity well, and they tend to underestimate total diversity. Ultimately, the nature of a collection, or how well the distribution represents that of the entire metacommunity, has the greatest impact on the accurateness of theta. The representative nature of a sample is affected by patchiness but can be minimized by sampling over the spatial and temporal extent of a metacommunity.