Model features

[Camilo Andres Sierra Fandiño]

Hi Chirs,

I am reading the ctmm r package paper to understand the theory of the ctmm. 

I found that the three main features to classify CTPS models are:

1) position autocorrelation

2) velocity autocorrelation

3) range residency

but I do not understand very well the second feature (velocity autocorrelation), can you give me some explanation ? can you give me a example ?

Thanks Chirs

Regards,

Camilo Sierra 

[Christen Fleming]

Hi Camilo,

Your velocity at one time can be correlated with your velocity at another time, if those times are not too far apart. Physically, this must be true at some timescale because the alternative implies infinite forces and accelerations. Behaviorally, this will be true for even longer timescales than what physics dictates because animals move to get from A to B and it takes some time to do that, during which the velocity will be highly correlated. Velocity autocorrelation always exists and the question is whether or not the data are finely sampled enough to detect it.

Here is some code to play with:

# models with the same asymptotic diffusion rate
BM <- ctmm(range=FALSE,tau=Inf,sigma=1)
IOU <- ctmm(range=FALSE,tau=c(Inf,1 %#% 'hr'),sigma=1)
# simulations with same sampling schedule
dt <- 1 %#% 'sec'
t <- seq(0,1 %#% 'hr',dt)
SIM.BM <- simulate(BM,t=t)
SIM.IOU <- simulate(IOU,t=t)
# plot
col <- rgb(1,0,0,0.1)
plot(SIM.BM,col=col)
plot(SIM.IOU,col=col)

and you can see that no matter how fine you make dt,  the model without velocity autocorrelation has no persistence of motion, but is fractal in its path.

Best,

Chris

[Camilo Andres Sierra Fandiño]

Hi Chris,

Many thanks for the reply and the explanation of my last question.   I following the workflow and to identify and visualized  my data I make the next scatterplot. 

The scatterplot help me see  the distribution of spatial point, but what I do not how identify outliers, migration events and other coarse features of the data. 

What is your opinion about this scatterplot ? 

What advice do you give me to interpret it ? 

Best, 

Camilo 

[Christen Fleming]

Hi Camilo,

There is a function outlie() in ctmm to help identify outliers. I don't see any obvious outliers from this.

Looking at this plot, I see two dens/nests where the points are clustered and two movement behaviors (a slower tortuous one at the clusters and a faster directed one elsewhere).

With the stationary models currently in the package, you would want to segment the data for both range and speed estimation. There are some convenience functions in ctmm to make this easy (e.g., lasso ), but currently there is a bug in RStudio where these functions may only work in the vanilla RGui.

Best,

Chris