Coming to the Eco-Stats '15 Conference? Would you like to know more about the speakers and their research before coming? We are compiling a reading list of suggested papers - one per speaker - and are holding a discussion group on Fridays 1-2pm to work through the list at UNSW (AGSM Courtyard). If you can't be there in person, you can contribute on Twitter using #ecostats15, Fridays at 1 (Sydney time)!
Sunday 27 September 2015
Otso Ovaskainen - Studying species interactions
The second paper in our reading group, for Friday (2nd October) at 1pm, is by Otso Ovaskainen from the University of Helsinki. Otso studies population dynamics and species co-occurrence in ecology with an interest in modelling metagenomics data.
So here are the main things we discussed: - It's a bit surprising that it took so long for people to develop models for this important problem. But it was technically very hard to do until maybe the mid 00's. - They sub-sampled the data quite heavily before analysis, and broke it up for separate models for each site, which seemed to waste a fair but of potentially useful information. This was probably for computational (takes too long otherwise) and technical (otherwise need to add random effects) reasons. - Things are moving quickly in this field and have progressed a long way even in the last five years: at ISEC last year, Otso presented about using latent variable models at multiple spatial scales to estimate and compare the nature of species interactions at multiple spatial scales, which is pretty cool.
Thanks for the summary! It does seem like things have advanced quite a bit since this paper was introduced. Otso is giving a short-course at IBS in Hobart as well, for anyone interested.
So here are the main things we discussed:
ReplyDelete- It's a bit surprising that it took so long for people to develop models for this important problem. But it was technically very hard to do until maybe the mid 00's.
- They sub-sampled the data quite heavily before analysis, and broke it up for separate models for each site, which seemed to waste a fair but of potentially useful information. This was probably for computational (takes too long otherwise) and technical (otherwise need to add random effects) reasons.
- Things are moving quickly in this field and have progressed a long way even in the last five years: at ISEC last year, Otso presented about using latent variable models at multiple spatial scales to estimate and compare the nature of species interactions at multiple spatial scales, which is pretty cool.
Thanks for the summary! It does seem like things have advanced quite a bit since this paper was introduced. Otso is giving a short-course at IBS in Hobart as well, for anyone interested.
DeleteI imagine this kind of approach would be especially difficult for presence-only data, particularly for things that move.
ReplyDelete