Integrating multiple field measurements in a Bayesian parallel regression framework to estimate Tasmanian devil age
Data files
Dec 27, 2025 version files 30.98 KB
Abstract
Understanding age structure in populations is foundational for ecological study. Yet ascertaining the age of individuals in the wild can be problematic. Methods typically rely on the prior determination of a relationship between age and some morphological measurement specific to the species or animal group under investigation. We present a novel method to incorporate multiple measures of animal size and development to estimate age, using Bayesian parallel regression to integrate multiple regression relationships into predictions. We apply the method to Tasmanian devils, a carnivorous marsupial that is threatened by a transmissible cancer. We estimate devil age based on measures of mass, head width, canine over-eruption and molar characteristics. This dataset includes data required to run and test code. See published article for full methodology.
Dataset DOI: 10.5061/dryad.05qfttfh8
Description of the data and file structure
This dataset constituted 2,448 capture events of 1,119 unique Tasmanian devils. Six were caught more than ten times, while 93 individuals were caught more than five times, and 653 individuals were captured only once. On each capture, a range of morphometric and dental records were collected, providing a suite of measurements that relate to growth and age.
Dataset is for use in the R Language and Environment for Statistical Computing.
Variables are described below:
dN - number of data records (2,448 captures)
dCount - number of individuals (1,119 devils)
dIndex - links each individual to their capture records for multiple captures
dCaptDate - capture dates in continuous time (days)
dMass - devil mass (kgs)
dSex - devil gender (coded as dSex == 1 for males, dSex == 2 for females)
dHead - devil head width (mm)
dTD - devil canine overeruption (mm)
dM2 - 2nd molar wear (Molar wear was scored from 0 to 5 in the field, where 0 indicates no damage, 1 the tip has been worn, 2 that some dentine is visible, 3 that the cusps are identical, 4 that just half or less of the tooth is present, and 5 that the tooth is worn flat to the gum. For the purposes of modelling, all scores were incremented by one (thus ranging 1-6). Any missing data were grouped using a seventh dummy score.)
dE4 - 4th molar eruption (molar eruption is scored from 0 to 3, such that 0 indicates the tooth is still below the gum, 1 indicates the first cusp is exposed, 2 that two cusps are exposed, and three that the tooth has fully erupted. For modelling, all scores were incremented by one (thus scores range from 1-4). Any missing data were grouped using a fifth dummy score.)
Files and variables
File: 20240828_Data_for_Kerlin_et_al_Integrating_multiple_field_measurements_to_predict_Tasmanian_devil_age.Rdata
Description: R data file containing data for running model as described in the paper
Code/software
Code has been provided in the Supplementary Info for the manuscript. All modelling was conducted using the JAGS program (Plummer 2003) and R 4.0.2 (R Core Team 2020).
Plummer, M. (2003) JAGS: a program for analysis of Bayesian graphical models using Gibbs sampling. Proceedings of the 3rd International Workshop on Distributed Statistical Computing, pp. 611. Vienna, Austria.
R Core Team (2020) R: A language and environment for statistical computing. R Foundation for Statistical Computing. Vienna, Austria.