Case study 2: von Bertalanffy growth with lizard size data

Load dependencies

# remotes::install_github("traitecoevo/hmde")
# install.packages(c("dplyr", "ggplot2"))

library(hmde)
library(dplyr)
library(ggplot2)

Overview

Our second demo introduces size-dependent growth based on the von Bertalanffy function \[\begin{equation}\label{eqn_vb_DE} f(Y(t); S_{max}, \beta) = \beta(S_{max} - Y(t)), \end{equation}\] where \(S_{max}\) is the asymptotic maximum size and \(\beta\) controls the growth rate. We have implemented the analytic solution \[\begin{equation}\label{eqn_vb_soln} Y(t) = S_{max} + (Y(0) - S_{max}) \exp(-t\beta) \end{equation}\] which is independent of age at the starting size \(Y(0)\) and instead uses the first size as the initial condition. The key behaviour of the von Bertalanffy model is a high growth rate at small sizes that declines linearly as the size approaches \(S_{max}\). This manifests as growth slowing as a creature matures with a hard finite limit on the eventual size. We restrict \(\beta\) and \(S_{max}\) to be positive. As a result the growth rate is non-negative.

Priors

The default priors for the constant top-level parameters in the single individual model are \[S_{max} \sim \log\mathcal{N}(\log(\max(y_{obs})), 2),\] \[\beta \sim \log\mathcal{N}(0, 2),\] \[0 <\sigma_e \sim Cauchy(0, 2).\]

For the multi-individual model the prior structure and default parameters are \[\mu_{S_{max}} \sim \mathcal{N}(\log(\max(y_{obs})), 2),\] \[0 < \sigma_{S_{max}} \sim Cauchy(0, 2),\] \[\mu_{\beta} \sim \mathcal{N}(0, 2),\] \[0 < \sigma_{\beta} \sim Cauchy(0, 2),\] \[0 < \sigma_{e} \sim Cauchy(0, 2).\] The max size parameter priors are always centred at the (transformed) maximum observed size. This is not changeable, but the standard deviation is. To see the name for the prior parameter run hmde_model. For example in the following we want to change the prior for \(S_{max}\) standard deviation (ind_max_size) in the individual model:

hmde_model("vb_single_ind")
#> $n_obs
#> NULL
#> 
#> $y_obs
#> NULL
#> 
#> $obs_index
#> NULL
#> 
#> $time
#> NULL
#> 
#> $y_bar
#> NULL
#> 
#> $prior_pars_ind_max_size_sd_only
#> [1] 2
#> 
#> $prior_pars_ind_growth_rate
#> [1] 0 2
#> 
#> $prior_pars_global_error_sigma
#> [1] 0 2
#> 
#> $model
#> [1] "vb_single_ind"
#> 
#> attr(,"class")
#> [1] "hmde_object"
#prior_pars_ind_max_size_sd_only is the argument name for the prior parameter

Visualise model

In the following code we plot an example of the growth function and the solution to get a feel for the behaviour.

#Analytic solution in function form
solution <- function(t, pars = list(y_0, beta, S_max)){
  return(
    pars$S_max + (y_0 - pars$S_max)*exp(-t * pars$beta)
  )
}

#Parameters
beta <- 0.35 #Growth rate
y_0 <- 1 #Starting size
S_max <- 20 #Asymptotic max size
time <- c(0,30) 
pars_list <- list(y_0 = y_0,
                  beta = beta,
                  S_max = S_max)
y_final <- solution(time[2], pars_list)

#Plot of growth function
ggplot() +
  xlim(y_0, y_final) +
  ylim(0, beta*(S_max-y_0)*1.1) +
  labs(x = "Y(t)", y = "f", title = "von Berralanffy growth") +
  theme_classic() +
  theme(axis.text=element_text(size=16),
        axis.title=element_text(size=18,face="bold")) +
  geom_function(fun=hmde_model_des("vb_single_ind"), 
                args=list(pars = list(S_max, beta)),
                colour="green4", linewidth=1,
                xlim=c(y_0, y_final))


#Size over time
ggplot() +
  geom_function(fun=solution, 
                args=list(pars = pars_list),
                colour="green4", linewidth=1,
                xlim=c(time)) +
  xlim(time) +
  ylim(0, y_final*1.05) +
  labs(x = "Time", y = "Y(t)", title = "von Bertalanffy growth") +
  theme_classic() +
  theme(axis.text=element_text(size=16),
        axis.title=element_text(size=18,face="bold"))

The von Bertalanffy model is commonly used in fishery management (Flinn and Midway 2021), but has also been used in reptile studies such as Edmonds et al. (2021) and Zhao et al. (2020).

Lizard size data

Our data is sourced from Kar, Nakagawa, and Noble (2023) which measured mass and snout-vent-length (SVL) of delicate skinks – – under experimental conditions to examine the effect of temperature on development. We are going to use the SVL metric for size.

We took a simple random sample without replacement of 50 individuals with at least 5 observations each. The von Bertalanffy model can be fit to shorter observation lengths, but fewer than 3 observations is not advised as there are two growth parameters per individual.

Implementation

The workflow for the second example is the same as the first, with the change in model name and data object.

lizard_vb_fit <- hmde_model("vb_multi_ind") |>
  hmde_assign_data(data = Lizard_Size_Data)  |>
  hmde_run(chains = 4, cores = 1, iter = 2000)
#> 
#> SAMPLING FOR MODEL 'vb_multi_ind' NOW (CHAIN 1).
#> Chain 1: 
#> Chain 1: Gradient evaluation took 0.000102 seconds
#> Chain 1: 1000 transitions using 10 leapfrog steps per transition would take 1.02 seconds.
#> Chain 1: Adjust your expectations accordingly!
#> Chain 1: 
#> Chain 1: 
#> Chain 1: Iteration:    1 / 2000 [  0%]  (Warmup)
#> Chain 1: Iteration:  200 / 2000 [ 10%]  (Warmup)
#> Chain 1: Iteration:  400 / 2000 [ 20%]  (Warmup)
#> Chain 1: Iteration:  600 / 2000 [ 30%]  (Warmup)
#> Chain 1: Iteration:  800 / 2000 [ 40%]  (Warmup)
#> Chain 1: Iteration: 1000 / 2000 [ 50%]  (Warmup)
#> Chain 1: Iteration: 1001 / 2000 [ 50%]  (Sampling)
#> Chain 1: Iteration: 1200 / 2000 [ 60%]  (Sampling)
#> Chain 1: Iteration: 1400 / 2000 [ 70%]  (Sampling)
#> Chain 1: Iteration: 1600 / 2000 [ 80%]  (Sampling)
#> Chain 1: Iteration: 1800 / 2000 [ 90%]  (Sampling)
#> Chain 1: Iteration: 2000 / 2000 [100%]  (Sampling)
#> Chain 1: 
#> Chain 1:  Elapsed Time: 6.642 seconds (Warm-up)
#> Chain 1:                2.511 seconds (Sampling)
#> Chain 1:                9.153 seconds (Total)
#> Chain 1: 
#> 
#> SAMPLING FOR MODEL 'vb_multi_ind' NOW (CHAIN 2).
#> Chain 2: 
#> Chain 2: Gradient evaluation took 0.000104 seconds
#> Chain 2: 1000 transitions using 10 leapfrog steps per transition would take 1.04 seconds.
#> Chain 2: Adjust your expectations accordingly!
#> Chain 2: 
#> Chain 2: 
#> Chain 2: Iteration:    1 / 2000 [  0%]  (Warmup)
#> Chain 2: Iteration:  200 / 2000 [ 10%]  (Warmup)
#> Chain 2: Iteration:  400 / 2000 [ 20%]  (Warmup)
#> Chain 2: Iteration:  600 / 2000 [ 30%]  (Warmup)
#> Chain 2: Iteration:  800 / 2000 [ 40%]  (Warmup)
#> Chain 2: Iteration: 1000 / 2000 [ 50%]  (Warmup)
#> Chain 2: Iteration: 1001 / 2000 [ 50%]  (Sampling)
#> Chain 2: Iteration: 1200 / 2000 [ 60%]  (Sampling)
#> Chain 2: Iteration: 1400 / 2000 [ 70%]  (Sampling)
#> Chain 2: Iteration: 1600 / 2000 [ 80%]  (Sampling)
#> Chain 2: Iteration: 1800 / 2000 [ 90%]  (Sampling)
#> Chain 2: Iteration: 2000 / 2000 [100%]  (Sampling)
#> Chain 2: 
#> Chain 2:  Elapsed Time: 17.996 seconds (Warm-up)
#> Chain 2:                3.827 seconds (Sampling)
#> Chain 2:                21.823 seconds (Total)
#> Chain 2: 
#> 
#> SAMPLING FOR MODEL 'vb_multi_ind' NOW (CHAIN 3).
#> Chain 3: 
#> Chain 3: Gradient evaluation took 9.3e-05 seconds
#> Chain 3: 1000 transitions using 10 leapfrog steps per transition would take 0.93 seconds.
#> Chain 3: Adjust your expectations accordingly!
#> Chain 3: 
#> Chain 3: 
#> Chain 3: Iteration:    1 / 2000 [  0%]  (Warmup)
#> Chain 3: Iteration:  200 / 2000 [ 10%]  (Warmup)
#> Chain 3: Iteration:  400 / 2000 [ 20%]  (Warmup)
#> Chain 3: Iteration:  600 / 2000 [ 30%]  (Warmup)
#> Chain 3: Iteration:  800 / 2000 [ 40%]  (Warmup)
#> Chain 3: Iteration: 1000 / 2000 [ 50%]  (Warmup)
#> Chain 3: Iteration: 1001 / 2000 [ 50%]  (Sampling)
#> Chain 3: Iteration: 1200 / 2000 [ 60%]  (Sampling)
#> Chain 3: Iteration: 1400 / 2000 [ 70%]  (Sampling)
#> Chain 3: Iteration: 1600 / 2000 [ 80%]  (Sampling)
#> Chain 3: Iteration: 1800 / 2000 [ 90%]  (Sampling)
#> Chain 3: Iteration: 2000 / 2000 [100%]  (Sampling)
#> Chain 3: 
#> Chain 3:  Elapsed Time: 12.192 seconds (Warm-up)
#> Chain 3:                6.087 seconds (Sampling)
#> Chain 3:                18.279 seconds (Total)
#> Chain 3: 
#> 
#> SAMPLING FOR MODEL 'vb_multi_ind' NOW (CHAIN 4).
#> Chain 4: 
#> Chain 4: Gradient evaluation took 9.5e-05 seconds
#> Chain 4: 1000 transitions using 10 leapfrog steps per transition would take 0.95 seconds.
#> Chain 4: Adjust your expectations accordingly!
#> Chain 4: 
#> Chain 4: 
#> Chain 4: Iteration:    1 / 2000 [  0%]  (Warmup)
#> Chain 4: Iteration:  200 / 2000 [ 10%]  (Warmup)
#> Chain 4: Iteration:  400 / 2000 [ 20%]  (Warmup)
#> Chain 4: Iteration:  600 / 2000 [ 30%]  (Warmup)
#> Chain 4: Iteration:  800 / 2000 [ 40%]  (Warmup)
#> Chain 4: Iteration: 1000 / 2000 [ 50%]  (Warmup)
#> Chain 4: Iteration: 1001 / 2000 [ 50%]  (Sampling)
#> Chain 4: Iteration: 1200 / 2000 [ 60%]  (Sampling)
#> Chain 4: Iteration: 1400 / 2000 [ 70%]  (Sampling)
#> Chain 4: Iteration: 1600 / 2000 [ 80%]  (Sampling)
#> Chain 4: Iteration: 1800 / 2000 [ 90%]  (Sampling)
#> Chain 4: Iteration: 2000 / 2000 [100%]  (Sampling)
#> Chain 4: 
#> Chain 4:  Elapsed Time: 12.321 seconds (Warm-up)
#> Chain 4:                6.187 seconds (Sampling)
#> Chain 4:                18.508 seconds (Total)
#> Chain 4:
#> Warning: There were 269 divergent transitions after warmup. See
#> https://mc-stan.org/misc/warnings.html#divergent-transitions-after-warmup
#> to find out why this is a problem and how to eliminate them.
#> Warning: There were 2 chains where the estimated Bayesian Fraction of Missing Information was low. See
#> https://mc-stan.org/misc/warnings.html#bfmi-low
#> Warning: Examine the pairs() plot to diagnose sampling problems
#> Warning: Bulk Effective Samples Size (ESS) is too low, indicating posterior means and medians may be unreliable.
#> Running the chains for more iterations may help. See
#> https://mc-stan.org/misc/warnings.html#bulk-ess
#> Warning: Tail Effective Samples Size (ESS) is too low, indicating posterior variances and tail quantiles may be unreliable.
#> Running the chains for more iterations may help. See
#> https://mc-stan.org/misc/warnings.html#tail-ess

lizard_estimates <- hmde_extract_estimates(fit = lizard_vb_fit,
                                           input_measurement_data = Lizard_Size_Data)

As before, we can compare the observed sizes over time to those predicted by the model.

measurement_data_transformed <- lizard_estimates$measurement_data %>%
  group_by(ind_id) %>%
  mutate(
    delta_y_obs = y_obs - lag(y_obs),
    obs_interval = time - lag(time),
    obs_growth_rate = delta_y_obs/obs_interval,
    delta_y_est = y_hat - lag(y_hat),
    est_growth_rate = delta_y_est/obs_interval
  ) %>%
  ungroup()


#Distributions of estimated growth and size
hist(measurement_data_transformed$y_hat, 
     main = "Estimated size distribution",
     xlab = "Size (cm)")

hist(measurement_data_transformed$delta_y_est, 
     main = "Estimated growth increments",
     xlab = "Growth increment (cm)")

hist(measurement_data_transformed$est_growth_rate, 
     main = "Estimated annualised growth rate distribution",
     xlab = "Growth rate (cm/yr)")


#Quantitative R^2
cor(measurement_data_transformed$y_obs, measurement_data_transformed$y_hat)^2
#> [1] 0.7463553
r_sq_est <- cor(lizard_estimates$measurement_data$y_obs,
                lizard_estimates$measurement_data$y_hat)^2
r_sq <- paste0("R^2 = ", 
               signif(r_sq_est,
                      digits = 3))

obs_scatter <- ggplot(data = lizard_estimates$measurement_data, 
       aes(x = y_obs, y = y_hat)) +
  geom_point(shape = 16, size = 1, colour = "green4") +
  xlab("Y obs.") +
  ylab("Y est.") +
  geom_abline(slope = 1, linetype = "dashed") +
  annotate("text", x = 25, y = 18, 
           label = r_sq) +
  theme_classic()

#Plots of size over time for a sample of 5 individuals
obs_est_ind <- hmde_plot_obs_est_inds(n_ind_to_plot = 5,
                       measurement_data = lizard_estimates$measurement_data) +
  theme(legend.position = "inside",
        legend.position.inside = c(0.8, 0.2))

We have two parameters at the individual level and are interested in both their separate distributions, and if we see evidence of a relationship between them. We can also use the individual parameter estimates and estimated sizes to plot the growth function pieces.

#1-dimensional parameter distributions
s_max_hist <- ggplot(lizard_estimates$individual_data, 
       aes(ind_max_size_mean)) +
  geom_histogram(bins = 10,
                 colour = "black",
                 fill = "lightblue") +
  labs(x="S_max estimate") +
  theme_classic()

beta_hist <- ggplot(lizard_estimates$individual_data, 
       aes(ind_growth_rate_mean)) +
  geom_histogram(bins = 10,
                 colour = "black",
                 fill = "lightblue") +
  labs(x="beta estimate") +
  theme_classic()

#2-dimensional parameter distribution
par_scatter <- ggplot(data = lizard_estimates$individual_data, 
       aes(x = ind_max_size_mean, y = ind_growth_rate_mean)) +
  geom_point(shape = 16, size = 1, colour = "green4") +
  xlab("Individual max sizes (mm)") +
  ylab("Individual growth rate parameters") +
  theme_classic()

#Correlation of parameters
cor(lizard_estimates$individual_data$ind_max_size_mean,
    lizard_estimates$individual_data$ind_growth_rate_mean,
    method = "spearman")
#> [1] 0.537383

#Plot function pieces over estimated sizes.
de_pieces <- hmde_plot_de_pieces(lizard_estimates)

At the hyper-parameter level for the whole population we have centre and spread parameters for the log-normal distributions of \(S_{max}\) and \(\beta\). As before, we can look at these as species-level features.

pars_CI_names <- c(
  "mean log max size",
  "mean max size in mm",
  "log max size standard deviation",
  "mean log growth par",
  "mean growth par mm/yr",
  "log growth par standard deviation"
)

#Vector that picks out which pars to be exponentiated
exp_vec <- c(FALSE, TRUE, FALSE, 
             FALSE, TRUE, FALSE)

#Print mean estimates and CIs
for(i in 1:nrow(lizard_estimates$population_data)){
  if(!exp_vec[i]){
    lizard_estimates$population_data$mean[i] 
    print(paste0("95% CI for ", 
                 pars_CI_names[i],
                 ": (",
                 lizard_estimates$population_data$CI_lower[i],
                 ", ",
                 lizard_estimates$population_data$CI_upper[i],
                 ")"))
  } else {
    exp(lizard_estimates$population_data$mean[i])
    print(paste0("95% CI for ",
                 pars_CI_names[i], 
                 ": (",
                 exp(lizard_estimates$population_data$CI_lower[i]),
                 ", ",
                 exp(lizard_estimates$population_data$CI_upper[i]),
                 ")"))
  }
}
#> [1] "95% CI for mean log max size: (3.17780521037458, 3.21995765765063)"
#> [1] "95% CI for mean max size in mm: (1.01746329235211, 1.05073762223325)"
#> [1] "95% CI for log max size standard deviation: (-4.20068883170187, -3.77451413388146)"
#> [1] "95% CI for mean log growth par: (0.0292990644585441, 0.23541251547073)"

References

Edmonds, Devin, Michael J Dreslik, Jeffrey E Lovich, Thomas P Wilson, and Carl H Ernst. 2021. “Growing as Slow as a Turtle: Unexpected Maturational Differences in a Small, Long-Lived Species.” PloS One 16 (11): e0259978.
Flinn, Shane A, and Stephen R Midway. 2021. “Trends in Growth Modeling in Fisheries Science.” Fishes 6 (1): 1.
Kar, Fonti, Shinichi Nakagawa, and Daniel W Noble. 2023. “Heritability and Developmental Plasticity of Growth in an Oviparous Lizard.” OSF. https://doi.org/10.17605/OSF.IO/HJKXD.
Zhao, Wei, Yangyang Zhao, Rui Guo, Yue Qi, Xiaoning Wang, and Na Li. 2020. “Age and Annual Growth Rate Cause Spatial Variation in Body Size in Phrynocephalus Przewalskii (Agamid).” Ecology and Evolution 10 (24): 14189–95.