Prior choice and data requirements of Bayesian multivariate mixed effects models fit to tag-recovery data: The need for power analyses
Data files
Feb 17, 2023 version files 220.78 MB
-
MALL_fahy_mij_core_1961_1996.txt
3.32 KB
-
MALL_fhy_mij_core_1961_1996.txt
3.32 KB
-
MALL_mahy_mij_core_1961_1996.txt
3.55 KB
-
MALL_mhy_mij_core_1961_1996.txt
3.49 KB
-
mij_intensive_14174.RDS
83.98 KB
-
mij_intermediate_14174.RDS
60.05 KB
-
mij_modest_14174.RDS
41.67 KB
-
mij_pulsed_14174.RDS
43.11 KB
-
MR-f6196-gamma.RDS
73.09 MB
-
MR-f6196-unif.RDS
73.77 MB
-
MR-f6196-wish3.RDS
73.68 MB
-
README_Dataset-PriorsDataRequirementsBayesianHierarchicalModels.md
6.83 KB
Aug 13, 2024 version files 220.79 MB
-
MALL_fahy_mij_core_1961_1996.txt
3.32 KB
-
MALL_fhy_mij_core_1961_1996.txt
3.32 KB
-
MALL_mahy_mij_core_1961_1996.txt
3.55 KB
-
MALL_mhy_mij_core_1961_1996.txt
3.49 KB
-
mij_intensive_14174.RDS
83.98 KB
-
mij_intermediate_14174.RDS
60.05 KB
-
mij_modest_14174.RDS
41.67 KB
-
mij_pulsed_14174.RDS
43.11 KB
-
MR-f6196-gamma.RDS
73.09 MB
-
MR-f6196-unif.RDS
73.77 MB
-
MR-f6196-wish3.RDS
73.68 MB
-
README_Dataset-PriorsDataRequirementsBayesianHierarchicalModels.md
6.83 KB
-
README.md
8.06 KB
Abstract
1. Recent empirical studies have quantified correlation between survival and recovery by estimating these parameters as correlated random effects with hierarchical Bayesian multivariate models fit to tag-recovery data. In these applications, increasingly negative correlation between survival and recovery has been interpreted as evidence for increasingly additive harvest mortality. The power of these hierarchal models to detect non-zero correlations has rarely been evaluated and these few studies have not focused on tag-recovery data, which is a common data type.
2. We assessed the power of multivariate hierarchical models to detect negative correlation between annual survival and recovery. Using three priors for multivariate normal distributions, we fit hierarchical effects models to a mallard (Anas platyrhychos) tag-recovery dataset and to simulated data with sample sizes corresponding to different levels of monitoring intensity. We also demonstrate more robust summary statistics for tag-recovery datasets than total individuals tagged.
3. Different priors lead to substantially different estimates of correlation from the mallard data. Our power analysis of simulated data indicated most prior distribution and sample size combinations could not estimate strongly negative correlation with useful precision or accuracy. Many correlation estimates spanned the available parameter space (–1,1) and underestimated the magnitude of negative correlation. Only one prior combined with our most intensive monitoring scenario provided reliable results. Underestimating the magnitude of correlation coincided with overestimating the variability of annual survival, but not annual recovery.
4. The inadequacy of prior distributions and sample size combinations previously assumed adequate for obtaining robust inference from tag-recovery data represents a concern in the application of Bayesian hierarchical models to tag-recovery data. Our analysis approach provides a means for examining prior influence and sample size on hierarchical models fit to capture-recapture data while emphasizing transferability of results between empirical and simulation studies.
README: Data for the article "Prior choice and data requirements of Bayesian multivariate hierarchical models fit to tag-recovery data: the need for power analyses"
Reference information
- File name: README_Dataset-PriorsDataRequirementsBayesianHierarchicalModels.md
- Authors: C.E. Deane (cdeane2@alaska.edu)
- Other contributors: L.G. Carlson, C.J. Cunningham, P. Doak, K. Kielland, G.A. Breed.
- Date of Issue: 2023-02-15
- Suggested Citations:
- Dataset citation: > Deane, C.E., L.G. Carlon, C.J. Cunningham, P. Doak, K. Kielland, and G.A. Breed. 2023. Data for "Prior choice and data requirements of Bayesian multivariate hierarchical models fit to tag-recovery data: the need for power analyses", Dryad, Dataset, https://doi.org/10.5061/dryad.hmgqnk9h6
- Software citation: > Deane, C.E., L.G. Carlon, C.J. Cunningham, P. Doak, K. Kielland, and G.A. Breed. 2023. Software for "Prior choice and data requirements of Bayesian multivariate hierarchical models fit to tag-recovery data: the need for power analyses", Zenodo, Software, https://doi.org/10.5281/zenodo.13275071
- Corresponding publication: > Deane, C.E., L.G. Carlon, C.J. Cunningham, P. Doak, K. Kielland, and G.A. Breed. 2023. Prior choice and data requirements of Bayesian multivariate hierarchical models fit to tag-recovery data: the need for power analyses. Ecology and Evolution. Accepted. DOI: 10.22541/au.166004819.97373134/v1
Contact information
- Primary contact
- Name: Cody E. Deane
- Affiliation: Department of Biology and Wildlife, P.O. Box 756100, Fairbanks, AK 99775, USA
- ORCID ID: https://orcid.org/0000-0002-4098-9154
- Email: cdeane2@alaska.edu
- Alternate email: codyedeane@gmail.com
- Alternative Contact
- Name: Greg Breed
- Affiliations: Department of Biology and Wildlife and Institute of Arctic BiologyP.O. Box 756100, Fairbanks, AK 99775, USA
- ORCID ID: https://orcid.org/0000-0002-7958-1877
- Email: gabreed@alaska.edu
File/Folder Details
Details for: MALL_fhy_mij_core_1961_1996.txt, MALL_fahy_mij_core_1961_1996.txt, MALL_mhy_mij_core_1961_1996.txt, MALL_mahy_mij_core_1961_1996.txt
=====================================
- Description: Mallard banding data and harvest recovery data summarized into m-array format for empirical analysis. See manuscript for geographic and temporal filters applied to mallard data and for details about the m-array format. Sex and age class codes are based at status when releases: fhy (female hatch year/juvenile), fahy (female after hatch-year/adult), mhy (female hatch year/juvenile), mahy (female after hatch-year/adult). These files are created by the R scripts MALL1BandingUpload.R, MALL2RecoveryUpload.R, MALL3CompleteMijArray.R from publically available banding and harvest recovery data maintained by the U.S. Geological Survey Bird Banding Lab (please email for additional info). These m-array data are used by the R script MALL4ExecuteModels.R for statistical analysis (available at Zenodo citation).
- Format: .txt
- Size: 4 KB
- Missing data codes: not applicable
Details for: MR-f6196-wish3.RDS
MR-f6196-unif.RDS
MR-f6196-gamma.RDS
============================
- Description: Estimates of annual survival and recovery for juvenile and adult female mallards between 1961-1996. These parameters were estimated from the m-arrays (described above) using the R script MALL4ExecuteModels.R; each .RDS file contines the posterior distributions for monitored parameters for our three models implemented with Wishart (-wish3.RDS) priors, Gamma priors (-gamma.RDS), or Uniform priors (-unif.RDS). The file MR-f6196-unif.RDS are the results on which our power analysis is based; these results are uploaded by the file P22SimRealsMVN.R to simulate vital rates for the power analysis included in the manuscript. These data objects are complex lists.
- Format: .RDS (R data structure)
- Size: about 72,000 KB (each)
- Missing data codes: not applicable
Details for: mij_modest_14174.RDS
mij_pulsed_14174.RDS
mij_intermediate_14174.RDS
mij_intensive_14174.RDS
============================
*Description: These are the datasets from which parameters for our power analysis were estimated. These objects are created from simulated data (P22PlotRealsMVN.R and P23SimCaptureHistories.R) and the sampling code included in the R scripts P24ModestConstant.R, P24ModestPulsed.R, P24Intermediate.R, P24Intensive.R, respectively. As provided, these R scripts upload these .RDS files and execute the Bayesian statistical models so the user does not need to run the simulation and sampling code. These data objects are three dimensional arrays, with rows and columns corresponding to a single m-array and each slice being one of 50 m-arrays for a given monitoring scenario.
- Format: .RDS (R data structure)
- Size: 40-85 KB (each)
- Missing data codes: not applicable
Sharing/Access information
============================
This is a section for linking to other ways to access the data, and for linking to sources the data is derived from, if any.
Code/Software for Deane et al. 2023. Notes from February 16, 2023
============================
We fit Bayesian models using Program JAGS with code implemented in Program R (and RStudio) along with the following R packages: coda, jagsUI, MASS, rjags, coda. We uploaded two series of R scripts to Zenodo. One series is for the empirical/mallard analysis (MALL---.R) and one series is for the power analysis (P---.R) the the prefix is followed by a sequential numbering scheme and short description (eg. The beginning of each script has information on what a script follows, requires, and proceeds. The first simulation step of our power analysis and our folder organization/directories for the power analysis are based on a seed value '14174' that is included at the top of scripts for the power analysis.
Describe any scripts, code, or notebooks (e.g., R, Python, Mathematica, MatLab) as well as the software versions (including loaded packages) that you used to run those files. If your repository contains more than one file whose relationship to other scripts is not obvious, provide information about the workflow that you used to run those scripts and notebooks.
- Version information: JAGS (Verion 4.3.0) Program R (Version 4.1.2, 2021-11-01) RStudio (Version 2021.09.1 Build 372) coda (Version 0.19.4) jagsUI (Version 1.5.2) MASS (Version 7.3.54) rjags (Version 4.12)
Code/Software for Deane et al. in prep (response to Riecke et al.). These notes are new August 8, 2024
============================
We fit Bayesian models using Program JAGS with code implemented in Program R (and RStudio) along with the following R packages: coda, jagsUI, MASS, rjags, coda. We uploaded a new series of R scripts to Zenodo. This series begins with "R" for response (R---.R) and integrates with the "P" series (P---.R) for the power analysis (described above). The R series includes a script to create a new JAGS model using Gamma(1,1) as the prior distributions for the standard deviations. Then the new code executes these models for the realized datasets from the simulation in Deane et al. (2023); mij_modest_14174.RDS, mij_intermediate_14174.RDS, mij_intensive_14174.RDS
- Version information: JAGS (Verion 4.3.0) Program R (Version 4.1.2, 2021-11-01) RStudio (Version 2021.09.1 Build 372) coda (Version 0.19.4) jagsUI (Version 1.5.2) MASS (Version 7.3.54) rjags (Version 4.12)
Revised August 8, 2024
END OF README
Usage notes
Instructions and metadata are provided at the top of each R script, including required R packages listed at the top of individual R scripts. Software specifications are included in the README file included in the data files.
This Dryad repository is for two manuscripts:
-
Deane, C. E., L. G. Carlson, C. J. Cunningham, P. Doak, K. Kielland, and G. A. Breed. 2023. Prior choice and data requirements of Bayesian multivariate hierarchical models fit to tag-recovery data: The need for power analyses. Ecology and Evolution 13:e9847. https://doi.org/10.1002/ece3.9847 Note: R scripts specific to the power analysis in this manuscript begin with "P"
- Deane, C. E., L. G. Carlson, C. J. Cunningham, P. Doak, K. Kielland, and G. A. Breed. In prep. Accurately estimating correlations between demographic parameters: a response to Riecke et al. (in press). Ecology and Evolution Note: New R scripts specific to our response begin with the letter "R"