User:Carl Boettiger/Notebook/Stochastic Population Dynamics/2010/05/21

{| 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"|

Puzzled over the two-step jump process results
The variance dynamics are slightly different if the continuous time Markov process allows events to contain two steps (such as a transition of an individual from one age class to another, or one patch to another, etc) as well as one step birth/death) transitions.  $$ V\tau_{ij}(X/V) $$ to represent the rate of transitions from state i to state j.  We change variables for our state space from the discrete count state variable X by dividing by the system size (volume) V, to obtain the continuous, mean-field approximation of the density p = X/V. Then:



\begin{align} & \frac{d}{dt} E(p) = b(p)-d(p) + \sum_{j \neq i} \tau_{ji}(p) - \tau_{ij}(p) =: f(p)\\ & \frac{d}{dt} M = \textrm{diag}( b(p) + d(p) )  - T + Df(p) M + M Df(p)^T \end{align} $$

where T is a symmetric matrix with off-diagonal elements i,j $$ \tau_{ij}(p) + \tau{ji}(p) $$ and diagonal elements chosen so that rows sum to zero. This approximation is implemented in noise_approx.R

Example: Crowley model

 * Carl Boettiger 15:05, 25 May 2010 (EDT): (some corrections to eqns)

(Note this is the original Crowley model where the c coefficients are equal. In the original model this represented competition for space by overgrowth and hence would more naturally be expressed as a two step process).

\begin{align} \dot x &= b_1 x (K - x - y)/K - d_1 x + c x y/K \\ \dot y &= b_2 y (K - x - y)/K - d_1 y - c x y/K \end{align} $$

Interpreting the cxy terms as transitions rather than independent birth and death events results in the same mean-field dynamics but differing fluctuations. For instance, the covariance dynamics are:

\frac{d}{dt} \textrm{Cov}(x,y) = \left(b_1 (1-2x-y)-d_1 +c y + b_2(1-x-2y)-d_2 -c x \right)\textrm{Cov}(x,y) - (b_1 x +c x )\sigma_y^2 -(b_2 y -c y)\sigma_x^2 - cxy $$ Whereas interpreted as single-step birth and death the cxy term at the end is absent.

The variance dynamics are similarly altered.

Numerical implementation
This individual-based simulation is implemented in metapop.c and callable from R interface from ind_based_models.R as metapop_ibm. The file metapop.R specifies the dynamics of birth, death, transition matrix and the Jacobian so that linear_noise_approx function from above can be used to calculate the moments. This file also runs the crolwey_ibm model (from crowley.c, interface in the ind_based.models.R) and the linear noise approximation to it from crowley.R to compare how the two-step process changes the noise dynamics.

This comparison can be run simply by loading the package and running metapop_example function.

For some reason the individual based-simulation of the two-step process seems to resemble that of the one-step process more, while the two-step linear noise approximation alternately underestimates (Fig 1) or overestimates (Fig 2) the variance. Looking for bugs in the implementation at the moment.


 * Carl Boettiger 21:43, 26 May 2010 (EDT): Bug found, see corrected graphs on Tuesday's (5/25) entry.






 * }