sim_fit error with rw walk model

[AdriGooden]

Hi there

I am trying to create pseudo-tracks for my animals and need to create CRW or RW models, as my SSM was originally done under an MP model. However I am getting these errors when running sim_fit on my random walk model. This data set comprises 6 animals that would not converge under the CRW model. The crw model ran through the sim_fit process with no issues. I am not sure why as there are no NAs in the data and the all animals converged or if it is appropriate to do two different ssm models when creating absences. I have attached the data file. I have tried simplifying my RW model to remove the vmax and altering my timestep (although this is the time step that created good tracks under my MP and CRW models)
nonconverge_sharks.csv

SSM code used

fitrw <-   fit_ssm(x = df2, 
          vmax = 4, 
           model = "rw", 
           time.step = 24, # this is the time step I have used for all my models with no issues
          control = ssm_control(verbose = 0)) 

This model fits fine as shown below

 id    ssm          converged pdHess pmodel
  <chr> <named list> <lgl>     <lgl>  <chr> 
1 179   <ssm [15]>   TRUE      TRUE   rw    
2 234   <ssm [15]>   TRUE      TRUE   rw    
3 25    <ssm [15]>   TRUE      TRUE   rw    
4 375   <ssm [15]>   TRUE      TRUE   rw    
5 40    <ssm [15]>   TRUE      TRUE   rw    
6 48    <ssm [15]>   TRUE      TRUE   rw

I tried two different methods for running the sim_fit function on the RW model (the first one ran no issues on my CRW)

null_fit <- 
  sim_fit(fitrw, 
          what = "predicted", 
          reps = 100)

Error in if (mu[2] < y_rng[1]) { : missing value where TRUE/FALSE

and

st <- sim_fit(fitrw[2,],
              what="predicted", 
              reps=5)    

Error in FUN(X[[i]], ...) : Implausible travel rates detected, check SSM fit for implausible locations

There doesn't appear to be any major issues with the locations and they align with others from these animals. I have not had any issues creating tracks from my crw model.

I appreciate your help with this.

Thanks

Adrienne

[AdriGooden]

library(aniMotum)
#> Warning: package 'aniMotum' was built under R version 4.3.2
#> Warning in checkTMBPackageVersion(): Package version inconsistency detected.
#> aniMotum was built with TMB version 1.9.10
#> Current TMB version is 1.9.9
#> Please re-install 'aniMotum' from source using install.packages('aniMotum', type = 'source') or remove your current TMB version and re-install using remotes::install_version('TMB', version = '1.9.10')
library(patchwork)
#> Warning: package 'patchwork' was built under R version 4.3.3
library(sf)
#> Warning: package 'sf' was built under R version 4.3.2
#> Linking to GEOS 3.11.2, GDAL 3.7.2, PROJ 9.3.0; sf_use_s2() is TRUE
library(tidyverse)
#> Warning: package 'tidyverse' was built under R version 4.3.2
#> Warning: package 'ggplot2' was built under R version 4.3.3
#> Warning: package 'readr' was built under R version 4.3.2
#> Warning: package 'dplyr' was built under R version 4.3.2
#> Warning: package 'forcats' was built under R version 4.3.2
#> Warning: package 'lubridate' was built under R version 4.3.3
library(purrr)
library(lubridate)
library(ggplot2)
library(readr)
setwd("C:/Users/adrie/OneDrive/Desktop/PhD/Chap 3A - MLRSF/SPOT DATA TAKE 2")
locsduplicate<-read.csv("locsduplicate.csv")

df2 <- subset(locsduplicate, (id %in% c(48,40,375,234,25,179))) # need to try and run a second model for the remainder of the sharks that wouldnt converge using the CRW
fitrw <-
  fit_ssm(x = df2, 
          vmax = 4, ## maximum speed of whale sharks (in m/s)
          model = "rw", ## Move persistence model
          time.step = 24, ## predict positions every 24 hours
          control = ssm_control(verbose = 0)) 
#> Warning in sqrt(as.numeric(object$diag.cov.random)): NaNs produced
#> Warning: Hessian was not positive-definite so some standard errors could not be calculated. Try simplifying the model by adding the following argument:
#>              map = list(psi = factor(NA))
print(fitrw, n= nrow(fitrw))
#> # A tibble: 6 × 5
#>   id    ssm          converged pdHess pmodel
#>   <chr> <named list> <lgl>     <lgl>  <chr> 
#> 1 179   <ssm [15]>   TRUE      TRUE   rw    
#> 2 234   <ssm [15]>   TRUE      TRUE   rw    
#> 3 25    <ssm [15]>   TRUE      TRUE   rw    
#> 4 375   <ssm [15]>   TRUE      TRUE   rw    
#> 5 40    <ssm [15]>   TRUE      TRUE   rw    
#> 6 48    <ssm [15]>   TRUE      TRUE   rw

###Both of these models produced different errors for the RW model###

null_fit <- 
  sim_fit(fitrw, ## the ssm model we want to base our null models on
          what = "predicted", ## component of the model to use
          reps = 100) ## number of replicated simulations per animal
#> Error in if (mu[2] < y_rng[1]) {: missing value where TRUE/FALSE needed

st <- sim_fit(fitrw[2,],
              what="predicted", 
              reps=5)    
#> Error in FUN(X[[i]], ...): Implausible travel rates detected, check SSM fit for implausible locations

library(terra)
#> Warning: package 'terra' was built under R version 4.3.2
#> terra 1.7.71
#> 
#> Attaching package: 'terra'
#> The following object is masked from 'package:tidyr':
#> 
#>     extract
#> The following object is masked from 'package:patchwork':
#> 
#>     area

load(system.file("extdata/grad.rda", package = "aniMotum"))
grad <- terra::unwrap(grad)
tibble(whaleshark_fit)
#> Error: object 'whaleshark_fit' not found
## Now lets rerun the simulation including the gradient object

null_fit2 <- 
  sim_fit(converged_tracks, 
          what = "predicted", 
          reps = 120, 
          grad = grad, 
          beta = c(-300,-300))
#> Error in eval(expr, envir, enclos): object 'converged_tracks' not found

plot(null_fit2, ncol = 6)
#> Error in h(simpleError(msg, call)): error in evaluating the argument 'x' in selecting a method for function 'plot': object 'null_fit2' not found

plot(null_fit, ncol = 3)
#> Error in h(simpleError(msg, call)): error in evaluating the argument 'x' in selecting a method for function 'plot': object 'null_fit' not found

^{Created on 2024-06-03 with reprex v2.1.0}

I have attached a reprex if that helps with answering the question

[ianjonsen]

A few things going on here. First, you need to heed the warnings at the start of this reprex - there's a package version mismatch between TMB and aniMotum. This is an annoying issue that I can't control, but you can solve it by first updating to the latest TMB pkg version and then reinstalling aniMotum from source so it gets build against the updated TMB pkg.

Second, the errors you reported in the first post are due to a couple of things. 1) Your shark tracks are extremely long-duration deployments with several very large data gaps. 2) In at least 1 of the SSM fits, you've got some implausible travel rates due to one or more outlier locations at either the start or the end of the track(s). Combined these mean that any location on the planet is plausible given the estimated travel rates of the sharks & the duration of the deployments (see figure). So I would have a look at the start and end of each track to remove any extreme locations - for example, perhaps you have included the tagging location at the start, or the pop-up location(s) at the end of the tracks - prior to fitting the SSM. You also might want to consider splitting the tracks at the large data gaps prior to fitting the SSM, although this won't be ideal for track simulation.

I've pushed a small update to aniMotum to deal with the first error you were getting:

Error in if (mu[2] < y_rng[1]) { : missing value where TRUE/FALSE

This shouldn't happen now.