chunshen1987/bayesian_analysis

This is a repository for bayesian analysis

Bayesian Analysis

This is a repository for bayesian analysis. This framework is built upon the hic-param-est package from Jonah Bernhard.

Installation

1. Install python3, with packages emcee, h5py, numpy, scikit-learn, scipy, matplotlib, corner. Use pip to install them if needed

2. If you don't have R, download R from here

3. Open an R Console instance by opening the R app or by typing R in the command line.

4. In the R console, type the command install.packages('lhs') and pick an appropriate download mirror if prompted. To ensure the package was properly installed, type library(lhs) in the R console. If that command runs without error, the package is installed. Close the R console by typing quit().

Test Run with a toy model

A toy model example can be run under the toy_model/ folder

1. The toy model

   y1 = A exp(-B x^2); y2 = C cosh(D x)

   A, B, C, D are the model parameters to fit.

2. The prior parameter range is defined in toy_model/ABCD.txt

3. Generating the training data sets

   python3 -m src.design -par toy_model/ABCD.txt -n 500 toy_model
   cd toy_model
   for i in `ls main`; do python3 toy_model.py -i main/$i; done
   mkdir model_results; mv run_* model_results
   

4. Generate the pseudo experimental data

   cd toy_model; python3 toy_model.py --exp

5. Run Bayesian Analysis

   The Bayesian analysis can be performed using the Jupyter Notebook Run_BayesianAnalysis.ipynb