User:Carl Boettiger/Notebook/Stochastic Population Dynamics/2010/05/10

From OpenWetWare
Jump to navigationJump to search
Stochastic Population Dynamics Main project page
Previous entry      Next entry

Alan Meeting


  1. Likelihood calculation for beetle model.
  2. Proposed model and calculations Linear parameter change for warning signal model. Results in a few test cases.
  3. Likelihood methods within SDE framework, an outline of methods. Subtle problems, good approximations, powerful approaches.


  1. LPA vs ELLPA model
  2. Defining state space in individual-based model vs stage based model, setting initial conditions? Conclusion: Means non-Markovian system, so likelihood isn't determined in one step approach. More lit review, see: Ellner [1], Perry, [2, 3, 4]
  3. Calculating likelihood while accounting for non-independence of the occupancy of each stage? Focus on one class? Conclusions: Choose Larvae.

Parameter fitting

On likelihood based model selection vs model based estimates of process:

  • Some cases maximum likelihood estimate agrees with statistical definition -- variance definition is mle of sigma parameter for indep gaussian random numbers, the arithmetic mean is the mle for lambda in the poisson, etc. Doesn't necessarily hold true in these time-series analyses, at least for numerical fitting. Would be better to estimate theta from the mean and search for other parameters.
  • requires more precise parameterization


  1. Ellner SP and Rees M. Integral projection models for species with complex demography. Am Nat. 2006 Mar;167(3):410-28. DOI:10.1086/499438 | PubMed ID:16673349 | HubMed [Ellner]
  2. De Valpine P. Stochastic development in biologically structured population models. Ecology. 2009 Oct;90(10):2889-901. DOI:10.1890/08-0703.1 | PubMed ID:19886497 | HubMed [Perry]
  3. Polansky L, de Valpine P, Lloyd-Smith JO, and Getz WM. Likelihood ridges and multimodality in population growth rate models. Ecology. 2009 Aug;90(8):2313-20. DOI:10.1890/08-1461.1 | PubMed ID:19739392 | HubMed [Perry2]
  4. de Valpine P. Shared challenges and common ground for Bayesian and classical analysis of hierarchical statistical models. Ecol Appl. 2009 Apr;19(3):584-8. DOI:10.1890/08-0562.1 | PubMed ID:19425421 | HubMed [Perry3]
All Medline abstracts: PubMed | HubMed

Research: Themes for Weds/Thurs workshop with Bob and Brett

Why is this all important?

  • Importance of space
  • Importance of environmental variation
  • Importance of intrinsic variation <- us

A recipe for trouble

  • Age or stage structure
  • Density dependence
  • demographic stochasticity

Reading / Misc

Open Access

  • Answering a question for a colleague about whether posting on arxiv was consistent with Elsevier's policies. (It is). Many Eslevier journals including TPB provide a watermarked authors copy of post-publication pdf explicitly for posting on the author's website and sharing with colleagues.

  • Fun stats: 90% of journals allow archiving preprints, many major funders (NIH, Wellcome) and require open archives, and the FRPAA bill now before congress would require it for almost all public funding. Very few publishers maintain the Ingelfinger rule of refusing to publish things that are archived before publication. Open archives are now the norm.

Open Education

  • This is how we dream a fantastic piece on the changing media of writing. part 2. From Richard Miller, at Rutgers.
  • AcademiX an exciting conference on the open future of learning.