Uncertainty and Sensitivity Analysis using GenStat

The Biometris GenStat Procedure Library contains a collection of procedures mainly written by members of Biometris of Wageningen UR. Uncertainty and regression-based sensitivity analysis of a deterministic model can be performed using procedures EDCONTINUOUS, GMULTIVARIATE, GUNITCUBE, RUNCERTAINTY and SUMMARIZE:

EDCONTINUOUS calculates equivalent deviates for continuous distributions
GMULTIVARIATE generates random numbers from multivariate normal or Student’s t distribution
GUNITCUBE generates random numbers from a distribution with marginal uniform distributions
RUNCERTAINTY calculates contributions of model inputs to the variance of a model output
SUMMARIZE prints summary statistics for variates

The manual, which can be downloaded on the right side of this page, contains theory, examples and descriptions of these procedures. For the actual use of the software presented, a moderate amount of experience with GenStat is required. But we hope that a large part of the manual will also constitute interesting reading for those unfamiliar with GenStat.

The procedures can be used for sampling from continuous multivariate distributions of model input. Model output corresponding to the generated input samples is generally calculated outside GenStat. Various procedures are available for the subsequent analysis of uncertainty or sensitivity of the model output.