Just enter your email
A peer-reviewed article of this Preprint also exists.
An interesting article, although I find the 'statistical analysis' (lines 146-159) section confusing. In general it would help a reader to follow it if the procedures were linked to the hypotheses rather than a description of the procedures. For example: "In order to to test whether <hypothesis> we fitted a GLM... I find this useful so I can quickly go back to the methods if I need a reminder when reading the results.
Also in the same section, from the text I don't understand the tests used as reported in lines 149 - 151. Surely the parasite richness is count data and therefore needs a Poisson or Negative Binomial (etc) family model? You say you compare between sites, so is site a factor in the model formula and each 'observation' a host bee? Also I don't understand what's meant by "a linear distribution", presumably you refer to the model's error structure, for which linear isn't a valid choice. If you've used a normal error structure which is the default in most software then this is wrong (providing that as I understand it the data are counts of parasite taxa in individual bees), personally with data like you describe I'd start with using a Poisson GLM and if the residuals were overdispersed try using a pseudo-poisson before moving onto a Negative Binomial as a last resort (its highly unlikely that PASW supports that as an option).
As a reader, I like to scribble down the hypotheses as I read a paper and them annotate them with the model formulae and particulars and them tick them off as they're supported by the results. This is difficult for your paper as the reader has to work out for their self which model goes with which hypothesis and the formulae are not always clear. That said, really interesting and valuable results, which from your figures look like they support your conclusions.