A fast implementation of the Expected Value of Perfect Parameter Information (EVPPI) for large Monte Carlo simulations

General Information

In this repository you can find 4 things:

• a Python/Numpy implementation in python
• a C implementation in c
• R bindings to the C implementation in r
• Python bindings to the C implementation using CFFI in python_cffi

The Expected Value of Perfect Parameter Information (EVPPI) is a concept from decision analysis (modeling decisions under uncertainty). It can be described as a measure for what a (rational) decision-maker would be willing to pay for zero uncertainty on a certain variable.

In general, the functions in this repository take in samples from a Monte Carlo model that predicts utility as a function of uncertain input parameters. Here, x denotes the values of the (uncertain) parameter inputs and y the resulting utility. More detailed documentation can be found in the respective packages.

Running the C implementation from R was found to be many times faster than existing R implementations, especially for a large number of Monte Carlo samples.

Version Information

First version. All 4 parts (C, Python, Python/CFFI and R) should be ready to be used in production.

Full Changelog: https://github.com/johanneskopton/evpi/commits/v1.0.0