Outline content for the workshop

Simulations and modelling (in R)

  Simulations are an important tool in study design for wildlife research. They allow you to explore different sample sizes and sampling strategies, and see the kind of data and the quality of the results you can expect.

Should you use pure random or systematic random sampling? Should you focus on a few study sites, or spread your effort thinly over many sites? Will you be able to answer your research question with the resources available? And, of course, are you clear about how the data will be analysed?

Modelling and R

To generate simulated data, we need models: "Statistical models are stories about how the data came to be."1 You will have to think through how the data you expect to record might be generated, which will depend on what is really happening and how you make your observations. Putting these ideas into a mathematical form to generate data is a good test of how well you understand your model.

The R programming language is ideal for building models and simulating data, as well as conducting most kinds of analysis. We'll be using lots of R during the workshop, and applicants are required to complete an R Skills Review before being accepted for the workshop.

Topics to be covered

We will begin with basic methods illustrated with widely applicable examples:

  • Modelling biological processes vs observation processes.
  • Sample size for a simple experiment: using loops to generate thousands of samples.
  • Assessing accuracy, bias, and confidence interval coverage; boxplots and bee-swarm plots.
  • Alternatives to using loops.
  • Using parallel processing to speed up simulations.
  • Speeding up MCMC analysis for simulations.
  • Generating data from different models and checking the efficacy of model selection methods (AIC and relatives).
  • Simulating different effort-allocation strategies - many visits to a few sites vs a few visits to many sites.

Simulation strategies vary widely depending on the specific scenarios investigated. During the second part of the workshop, participants will work in groups on specific topics (eg, occupancy estimation, mark-recapture for density or survival, species diversity).

Who should attend?

Participants should be wildlife biologists doing quantitative work and who are keen to understand modern approaches to modelling and study design.

Competence in using R software is essential for this workshop, and applicants are required to complete an R Skills Review before being accepted.

Please note that this is NOT a basic statistics workshop: we expect people to be already familiar with the analysis methods we'll be using. If you want a stats workshop come to a Boot Camp!

All participants should come with a laptop computer with R and a spreadsheet package installed.

Back to BCSS Home Page

1. Dave Harris, http://www.noamross.net/blog/2013/6/17/harrisbbmle.html


Page updated 31 August 2018 by Mike Meredith