Skip to content
Bayesian modelling tutorial based on rethinking + greta
R
Branch: master
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
.gitignore general update Apr 2, 2019
README.md Minor correction of README.md May 2, 2019
data.xlsx
dbinomSuccessful.R
greta.yml
sessionInfo Change dzipois to dpois in laid eggs Apr 27, 2019
zipoisEggs.R

README.md

Bayesian modelling tutorial

Instructions

  1. Install rethinking, greta and their dependencies TensorFlow and RStan

The rethinking package requires rstan which might require you somewhat special compiling configurations. If you have trouble with the C++ compiler try adding the line CXX14 = g++ -std=c++1y -Wno-unused-variable -Wno-unused-function -fPIC to the file ~/.R/Makevars (create if necessary).

The greta in turn, requires the Python modules tensorflow and tensorflow_probability. To install those you can either:

  • i) use the following code in a fresh terminal,
conda create --name greta
conda activate greta
conda install tensorflow==1.12.0
conda install tensorflow-probability
conda update --all

to create a Conda environment called greta, or

  • ii) use the greta.yml file to copy my greta Conda enviroment,
conda env create -f greta.yml

I propose using CPU-only TensorFlow within Conda environments, but feel free to use alternatives at your own discretion. Quick side note: to create a .yml file you can use the command conda env export --no-builds > greta.yml.

  1. Install all packages listed on top of the two R scripts zipoisEggs.R and dbinomSuccessful.R. All of them are available on CRAN.

  2. Run either or both scripts - the first implements a rethinking-based zero-inflated Poisson regression on the number of fledglings, and a Poisson regression on the number of laid eggs, whereas the second implements a greta-based logistic regression on whether a female has successfully produced fledglings or not.

Acknowledgements

The dataset used in this work pertains to the publication:

I want to thank the author Christina Riehl for all insighful discussions and suggestions. I also want to thank Nick Golding and the community at the gretaforum for their kind contributions.

Enjoy, all feedback is welcome!

Francisco

You can’t perform that action at this time.