R-main/README.md

In benmoran11/rubias: Bayesian Inference from the Conditional Genetic Stock Identification Model

The files included in folder /rubias/R-main/ allow the reproduction of all data, statistics, and graphs presented in the publication ________________. They also include numerous graphs for further visualization of our model validation data, which were excluded from the publication for brevity.

cross-validation.R includes the first two methods of validation listed in the publication: reproduction of the Leave-One-Out Cross-Validation in Hasselman et al. (2015) with and without parametric bootstrap correction, and the application of Monte Carlo Cross-Validation to the same dataset. coalescent_sim.R documents the third validation method, the creation of simulated genotypes from the coalescent, followed by Leave-One-Out Cross-Validation.

Sourcing these independent files will create (among other files from intermediate steps) two data.frames for each validation method. One, ( )_rho_data, contains true and estimated reporting unit proportions (rho) created during the corresponding cross-validation. The second, ( )_rho_dev contains summary statistics describing the deviation of each estimation method from its corresponding true rho value. The validation is identified by the prefix, with coal, Hass, and mc representing coalescent, Leave-One-Out, and Monte Carlo cross-validations, respectively. For quick access, these output data are stored within the main rubias folder as cjfas_data.RData.

Sourcing will also create a multitude of rough graphs which visualize this data. .lg graphs plot the true vs. estimated rho value a la Figure 3, while .mse, .m.bias, and .prop.bias show mean squared error, mean bias, and mean proportional bias for each reporting unit, with and without parametric bootstrapping. Prefixes c, h, and mc represent coalescent, Leave-One-Out, and Monte Carlo cross-validations.

The remaining two files of the R-main folder use the data created by the previous two; however, they begin by loading cjfas_data.RData, and so can be run independent of the original simulation scripts. proportion_sig_test includes the statistics and rough figures documenting the relationship between the residual for any given simulation, $\tilde{\rho}_r - \rho^\mathrm{sim}_r$, and the structural metric $\frac{N_C}{P} - \rho^\mathrm{sim}_r$. cjfas_graphs.R contains the exact scripts used to generate the ggplot2 figures found in __________.

benmoran11/rubias documentation built on Feb. 1, 2024, 10:52 p.m.