Workshop Blogging II: Doing REAL work

(Part 1 on Monday)
This morning we got down to work. I helped Elizabeth on her phenology. Her idea is to use the numbers of butterflies counted on a site over a year to model how abundance changes over the season, i.e. to find the flight period(s). Some species only have one flight period in a summer, others have two (i.e. two generations per year). And these may change with latitude.

The first problem was finding some data. The UK data hasn’t been extracted yet, but she does have some US data. Only she didn’t have t on her computer. Bizarrely, I did (we had been playing with more complex models last year). So, we spent an hour or so writing BUGS code to fit a simple Gaussian curve to the abundances. The code is fairly easy, but we intend to fit a hierarchical model, i.e. each curve for a site has a set of parameters, and the distribution of these parameters between sites and years can then be modelled (e.g. looking at the effect of latitude on them). Once we had the code, and had debugged it, we tried to run it. BUGS simulates the distribution of the parameters, using a cunning scheme called MCMC. This creates a time series which will eventually find itself in a steady state that is the parameter distribution. The theory works, but it can take some time to reach that state. For some initial values, we found that the parameters would run off to silly values, and BUGS then hangs as it tries to calculate values like exp(50). The present state is that we need t play around a bit to stop this (work out good initial values, better priors etc.). We also need the UK data.
Now we’re all back together to go over what had been done in the last workshop – some simulation work on the effect of evolution of dispersal rate on range expansion. The take-home message is that it can affect the spread. So, if we want to model range expansion, we have to be concerned that if we use current estimates of dispersal rate to predict future dispersal as a response to climate change, we might underestimate the rate of expansion. In other words, this makes life more difficult for us.
Ah, lunchtime.