User:Carl Boettiger/Notebook/Stochastic Population Dynamics/2010/05/24

{| width="800"
 * style="background-color: #EEE"|[[Image:owwnotebook_icon.png|128px]] Stochastic Population Dynamics
 * style="background-color: #F2F2F2" align="center"|  |Main project page
 * style="background-color: #F2F2F2" align="center"|  |Main project page


 * colspan="2"|
 * colspan="2"|

Model roundup

 * 1) metapop  Two populations with one-step transitions.  IBM in C code, R interface to IBM and linear noise approximation.
 * 2) crowley One-step transitions version of metapop. R interface to IBM and linear noise.
 * 3) beetles Two-step transitions with exponential waiting times in age classes. R interface completed in ind_based_models.R
 * 4) hastings Simple two-step model. no R interfaces yet.
 * 5) tribolium Non-markov aging process. R interface completed in simulate.R.  includes likelihood inference as well.


 * Individual-based models, implemented in C using my gillespie library, interfaced in R in ind_based_models.R. Could probably benefit from C++ class structure, as the codes involve more written repetition than necessary, and renaming similar functions to avoid name conflicts in the object files for the R package.  All individual simulations now include parallelization of replicates the C level via the gillespie library, which seems to be working correctly even when called from R.


 * R files with model names include individual based simulation and the approximation, plotted in comparison.


 * Beetles simple noise simulation likewise does not agree with the approximation, still troubleshooting the approximation.



Other exploration
Revisit Phil Holmes notes on free-response protocol / information theory and Neyman-Pearson theorem.

Meetings

 * Schreiber Lab meeting, discussed articles 4 and 5 below.
 * Discussion with Graham, see phylogenetics notebook.
 * GTC meeting, more on toolbox, interesting stats discussion afterwards, neg binomial as number of failures before n successes, claim that this almost always outperforms poisson models (in information criteria sense) on ecological datasets -- nothing is ever uncorrelated...

Reading
A couple good articles today, see Mendeley for details.

1. Goel AK, Vattam SS, Rugaber S, et al. Learning Functional and Causal Abstractions of Classroom Aquaria The SBF Theory of Understanding of Complex Systems ACT : Interactive Construction of SBF Models. In Review. 2010.

2. Crutchfield JP, Whalen S. Structural Drift : The Population Dynamics of Sequential Learning. In review. 2010:1-14.

3. Raghib M, Hill Na, Dieckmann U. A multiscale maximum entropy moment closure for locally regulated space-time point process models of population dynamics. Journal of mathematical biology. 2010. Available at: http://www.ncbi.nlm.nih.gov/pubmed/20446087.

4. Lindegren M, Möllmann C, Nielsen A, et al. Ecological forecasting under climate change: the case of Baltic cod. Proceedings. Biological sciences / The Royal Society. 2010. Available at: http://www.ncbi.nlm.nih.gov/pubmed/20236982.

5. Berkley Ha, Kendall BE, Mitarai S, Siegel Da. Turbulent dispersal promotes species coexistence. Ecology Letters. 2010;13(3):360-371. Available at: http://blackwell-synergy.com/doi/abs/10.1111/j.1461-0248.2009.01427.x.

Misc Code
Apparently R doesn't have a convenient generalized eigenvalue solver (for case of a matrix on either side). Wrote quick interface to the gsl lapack routines. Github's gists seem an ideal way to store and share such code snippets, so here it is embedded. 


 * }