Replies: 1 comment 1 reply
-
|
That error is due to some of the individuals having steps of zero length. You therefore must include zero-mass parameters. See the {momentuHMM} vignette. If you want centers of attraction or centroids to affect the movement of the individuals, then this behavior must be specified via the |
Beta Was this translation helpful? Give feedback.
-
|
Hello @bmcclintock Thank you very much for responding so quickly. I have been working on it during this time. Now I'm stuck again. I hope you can help me, but if not, thank you anyway. You have put a lot of effort into this package and all the information about it. Let me tell you a little bit... First, I tried to fit the trajectories to a recharge dynamics model like in the buffalo example in the vignette, but I found that the animal movement was not biased towards the water points (probably because the time interval between locations is too large to detect it). Since I need the simulated animals to visit the water points, I thought it would be best to simulate a Hierarchical HMM with two levels: The pseudo-matrix I want the I feel like I'm heading in the right direction, but I'm getting errors related to parameter input. Additionally, I would like to know if I could obtain average speeds and some measure of angles for each state from the Regarding the number of observations per level in Here is the code and the data (in Google Drive), which includes the crawled trajectories and the spatial covariates. Thank you very much for the help. # 1. Data object for momentuHMM modeling
spatialCovs <- list(wp_intercept = r_stack$intercept,
wp_near = r_stack$near_wp,
wp_dist = wp_dist,
D.x = dist_scaled$rast.grad.x,
D.y = dist_scaled$rast.grad.y,
landcover = landco_rast)
hmmData <- prepData(crw_traj, spatialCovs = spatialCovs, altCoordNames = "mu")
head(hmmData)
# 2. Fitting the model to real trajectories
DM <- list(mu=matrix(c("mu.x_tm1","crw(mu.x_tm1,lag=1)","D.x", 0,0,0,0,
"mu.x_tm1", 0,"D.x","crw(mu.x_tm1,lag=1)",0,0,0,
"mu.y_tm1","crw(mu.y_tm1,lag=1)","D.y", 0,0,0,0,
"mu.y_tm1", 0,"D.y","crw(mu.y_tm1,lag=1)",0,0,0,
0, 0, 0, 0,1,0,0,
0, 0, 0, 0,0,1,0,
0, 0, 0, 0,0,0,1,
0, 0, 0, 0,0,0,1,
0, 0, 0, 0,1,0,0,
0, 0, 0, 0,0,1,0
),2*5,7,byrow=TRUE,
dimnames = list(c(paste0("mean.x_",1:2),paste0("mean.y_",1:2),
paste0("sigma.x_",1:2),paste0("sigma.xy_",1:2),
paste0("sigma.y_",1:2)),
c("mu_tm1","crw_1","D","crw_2","sigma_1","sigma_2",
"sigma.xy"))))
Par0 <- list(mu=c(1,0.5,0,0.1,log(50000),log(50000000),0))
names(Par0$mu) <- c("mu.x_tm1_1","crw(mu.x_tm1,lag=1)_1","D.x_1","crw(mu.x_tm1,lag=1)_2",
"sigma.x_1","sigma.x_2","sigma.xy_1")
fixPar=list(mu=c(1,NA,0,NA,NA,NA,0),
delta = c(0.1,0.9),
beta=matrix(c(NA,NA,
NA,NA,
NA,NA,
NA,NA,
0,0),5,byrow = TRUE))
hmmFit <- fitHMM(hmmData,
nbStates = 2,
dist = list(mu = "rw_mvnorm2"),
DM = DM,
estAngleMean = list(angle=FALSE),
Par0 = Par0,
fixPar = fixPar,
formula = ~ 1 + landcover +wp_near,
stateNames = c('foraging', 'transit'), fit = TRUE,
mvnCoords="mu"); hmmFit
# plot(hmmFit)
# 3. Hier simulation
# 3.1. Hier states
hierStates <- data.tree::Node$new("States")
hierStates$AddChild("Hydrated")
hierStates$Hydrated$AddChild("Foraging", state=1)
hierStates$Hydrated$AddChild("Transit", state=2)
hierStates$AddChild(name="Thirsty")
hierStates$Thirsty$AddChild("Drinking", state=3)
hierStates$Thirsty$AddChild("Seeking", state=4)
# plot(hierStates)
# 3.2. Hier distribution
hierDist <- data.tree::Node$new("dist")
hierDist$AddChild("level1")
hierDist$AddChild("level2")
hierDist$level2$AddChild("mu", dist = "rw_mvnorm2")
# 3.3. Hier formula
hierFormula <- data.tree::Node$new("Formula")
hierFormula$AddChild("level1", formula=~recharge(g0=~1,theta=~wp_near))
hierFormula$AddChild("level2", formula=~1 + landcover + wp_near)
# 3.4. Hier beta
hierBeta <- data.tree::Node$new("Beta")
hierBeta$AddChild("level1",
beta=matrix(c(0,-1,0,-1),2,2),g0=-0.3,theta=c(-0.25,5))
hierBeta$AddChild("level2")
hierBeta$level2$AddChild("Hydrated",
beta=getPar(hmmFit)$beta)
hierBeta$level2$AddChild("Thirsty",
beta=matrix(c(-100,-100,
-100,-100,
-100,-100,
-100,-100,
-100, 100),
nrow=5,
ncol=2, byrow=TRUE))
# 3.5. Hier delta
hierDelta <- data.tree::Node$new("Delta")
hierDelta$AddChild("level1",delta=matrix(c(0.5, 0.5),1))
hierDelta$AddChild("level2")
hierDelta$level2$AddChild("Hydrated",delta=matrix(getPar0(hmmFit)$delta, 1))
hierDelta$level2$AddChild("Thirsty",delta=matrix(c(0, 1),1))
# 3.6. DM
DM <- list(mu=matrix(c("mu.x_tm1","crw(mu.x_tm1,lag=1)","D.x", 0, 0, 0,0,0,0,0,
"mu.x_tm1", 0,"D.x", 0,"crw(mu.x_tm1,lag=1)", 0,0,0,0,0,
"mu.y_tm1", 0, 0,"D.x", 0,"crw(mu.x_tm1,lag=1)",0,0,0,0,
"mu.y_tm1", 0, 0,"D.x","crw(mu.y_tm1,lag=1)", 0,0,0,0,0,
"mu.x_tm1","crw(mu.x_tm1,lag=1)","D.y", 0, 0, 0,0,0,0,0,
"mu.x_tm1", 0,"D.y", 0,"crw(mu.y_tm1,lag=1)", 0,0,0,0,0,
"mu.y_tm1", 0, 0,"D.y", 0,"crw(mu.y_tm1,lag=1)",0,0,0,0,
"mu.y_tm1", 0, 0,"D.y","crw(mu.y_tm1,lag=1)", 0,0,0,0,0,
0, 0, 0, 0, 0, 0,1,0,0,0,
0, 0, 0, 0, 0, 0,0,1,0,0,
0, 0, 0, 0, 0, 0,0,0,1,0,
0, 0, 0, 0, 0, 0,0,1,0,0,
0, 0, 0, 0, 0, 0,0,0,0,1,
0, 0, 0, 0, 0, 0,0,0,0,1,
0, 0, 0, 0, 0, 0,0,0,0,1,
0, 0, 0, 0, 0, 0,0,0,0,1,
0, 0, 0, 0, 0, 0,1,0,0,0,
0, 0, 0, 0, 0, 0,0,1,0,0,
0, 0, 0, 0, 0, 0,0,0,1,0,
0, 0, 0, 0, 0, 0,0,1,0,0
),4*5,10,byrow=TRUE,
dimnames = list(c(paste0("mean.x_",1:4),paste0("mean.y_",1:4),paste0("sigma.x_",1:4),
paste0("sigma.xy_",1:4),paste0("sigma.y_",1:4)),
c("mu_tm1","crw_1","D_12", "D_34","crw_24","crw_3","sigma_1","sigma_24",
"sigma_3","sigma.xy"))))
# 3.7. Par
Par0 <- getPar0(hmmFit)
Par <- list(mu=c(Par0$Par$mu[1:2], D_12 = 0, D_34=-100, crw_24=Par0$Par$mu[[4]], crw_3=0.9, Par0$Par$mu[5],
sigma_24=Par0$Par$mu[[6]],sigma_3=log(1),Par0$Par$mu[7]))
# 3.8. Initial positions
posits.df <- split(matrix(c(723904,723589,726974,723929,725284,4097746,4096536,4095861,4097626,4100256),
nrow = 5, ncol = 2), 1:5)
# 3.9. Simulation
obsPerLevel <- data.tree::Node$new("simHierData")
obsPerLevel$AddChild("level1",obs=500)
obsPerLevel$AddChild("level2",obs=500)
hmmSim <- simHierData(nbAnimals = 5, hierStates = hierStates, hierDist = hierDist,
Par = Par, hierBeta = hierBeta, hierDelta = hierDelta,
hierFormula = hierFormula, spatialCovs = spatialCovs, DM = DM,
states = TRUE, initialPosition = posits.df, mvnCoords = "mu",
obsPerLevel = obsPerLevel) |
Beta Was this translation helpful? Give feedback.
-
Hi Britt and Théo. First of all, congratulations and thank you for the R package. I find it very useful and with great potential for telemetry analysis.
In my case, I want to do something more focused on simulation. I don't need very realistic trajectories, but I do want the simulated animals to visit water points in a similar frequency to how they would in reality. I am fitting the fitHMM model to real trajectories from marked individuals, using water points as activity centers and habitat type (forest, shrubland, grassland, etc.) as spatialCov. My problem is that this treats the water points as independent, so those that are far from the are not considered. But I want the simulated individuals to be attracted to the water points nearby, because the animals were marked in a specific area of the park, but I want to simulate trajectories in a greater extent, including other water points.
I have been looking at other issues and I have seen that perhaps this could be done with centroids instead of centers, in which case for each location and time the centroid is the water point which is closest to the animal. But I don't know how I could do this.
Another idea is to take the average attraction effect that the water points in the area where the original trajectory is located have, and use it as a value for all the centers in the simulation. But I don't know how to access the effect that each water point has on the model, and I think it would be less precise than the previous option.
On the other hand, when I model with one or two individuals, I have no problem, but when I include several (of the same species), I get the following error. I understand that it is a convergence problem because the variability between individuals is very high, but I am not sure about it.
Thank you very much for your time and help.
Best regards,
César.
Beta Was this translation helpful? Give feedback.
All reactions