Dakota Reference Manual  Version 6.4
Large-Scale Engineering Optimization and Uncertainty Analysis
 All Pages
drop_tolerance


Suppresses output of sensitivity indices with values lower than this tolerance

Specification

Alias: none

Argument(s): REAL

Default: All VBD indices displayed

Description

The drop_tolerance keyword allows the user to specify a value below which sensitivity indices generated by variance_based_decomp are not displayed.

Default Behavior

By default, all sensitivity indices generated by variance_based_decomp are displayed.

Usage Tips

For polynomial_chaos, which outputs main, interaction, and total effects by default, the univariate_effects may be a more appropriate option. It allows suppression of the interaction effects since the output volume of these results can be prohibitive for high dimensional problems. Similar to suppression of these interactions is the covariance control, which can be selected to be diagonal_covariance or full_covariance, with the former supporting suppression of the off-diagonal covariance terms (to save compute and memory resources and reduce output volume).

Examples

method,
  sampling
    sample_type lhs
    samples = 100
    variance_based_decomp
    drop_tolerance = 0.001