gsa-module

gsa-module is a Python3 package implementing several global sensitivity analysis methods for computer/simulation experiments. The implementation is based on a black-box approach where the computer model (or any generic function) is externally implemented to the module itself. The module accepts the model outputs and the design of experiment (optional, only for certain methods) and compute the associated sensitivity measures. The package also includes routines to generate normalized design of experiment file to be used in the simulation experiment based on several algorithms (such as simple random sampling or latin hypercube) as well as simple routines to post-processed multivariate raw code output such as its maximum, minimum, or average.

The general calculation flow chart involved in using the gsa-module can be seen in the figure below.

Main Features (v0.9.0)

• Capability to generate design of computer experiments using 4 different methods: simple random sampling (srs), latin hypercube sampling (lhs), sobol' sequence, and optimized latin hypercube using either command line interface gsa_create_sample or the module API via import gsa_module
• Sobol' quasi-random number sequence generator is natively implemented in Python3 based on C++ implementation of Joe and Kuo (2008).
• Randomization of the Sobol' quasi-random number using random shift procedure
• Optimization of the latin hypercube design is done via evolutionary stochastic algorithm (ESE)
• Generation of separate test points based on a given design using Hammersley quasi-random sequence
• Capability to generate design of computer experiments for screening analysis (One-at-a-time design), based on the trajectory design (original Morris) and radial design (Saltelli et al.)
• Capability to compute the statistics of elementary effects, standardized or otherwise both for trajectory and radial designs. The statistics (mean, mean of absolute, and standard deviation) are used as the basis of parameter importance ranking.
• Capability to estimate the first-order (main effect) Sobol' sensitivity indices using two different estimators (Saltelli and Janon).
• Capability to estimate the total effect Sobol' sensitivity indices using two different estimators (Sobol-Homma and Jansen).
• All estimated quantities are equipped with their bootstrap samples

Complete log of changes can be found in CHANGELOG.

Requirements

The module was developed and tested using the Anaconda Python distribution of Python v3.5. No additional package except the base installation of the distribution is required.

Installation

gsa-module is hosted on BitBucket.

After cloning the source:

git clone [email protected]:lrs-uq/gsa-module.git

the installation can be done easily from the local source directory:

pip install -e .

This will make the following available in the path:

• The python module gsa_module
• The executable gsa_create_sample
• The executable create_validset
• The executable gsa_morris_generate
• The executable gsa_morris_analyze

Documentation

Documentation for gsa-module is an on-going process. The current version can be found in the /docs folder and can be built (given that sphinx has been installed) with the make command:

make html

to build the html version of the documentation. Note that the html documentation used rtd-theme which can be installed via pip:

pip install sphinx-rtd-theme

The index file can then be found in:

./docs/build/html/index.html

The current version of the documentation is also hosted on readthedocs

License

The project is licensed under the MIT License.