I am working with several datasets coming from studies located along a latitudinal gradient and would like to relate alpha diversity (i.e. species richness) to latitude (and other abiotic variables). So what would be the best approach to correct species richness values taking into account the different sampling effort (i.e. total number of traps used in each study) in a GLM? Would estimating species richness using extrapolation (i.e. iNEXT) for a similar value of sample size or sample coverage be the best alternative? Thanks.
You could use sampling effort as a weighting variable.
If this is an insect study then you have other problems. Estimated diversity will be a function of the number of traps, the frequency that traps are deployed, the type of trap, the date samples were collected, and the bait (if any). Elevation could also be an issue. Date is important as the date, latitude, and elevation all interact to determine the growing degree days accumulated at any one sample date. Unless samples are from a long term project (http://mothscount.org/text/34/rothamsted_insect_survey.html) there is no real way to correct for most of these issues.
Hi Pedro,
you can use the sampling effort as an offset term in glm's. In this way you can keep your response variable as it is (without transformations) and keep using the same distribution for the errors (poisson or neg. binomial, I assume). Keep in mind that you'll probably have to use the log version of the sampling effort, since the default link function for poisson and nb is log. For example: glm(y ~ x + z + offset(log(number_of_traps)), family = "poisson", data).
Cheers!
Pedro Giovâni da Silva ,
If you don't have complete data on the number of traps, sampling dates or other detailed information, your approach is the best: extrapolate or interpolate to achieve a standard sample coverage in iNext, and use this standardized richness value.
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Sampling Design
- Sample Size