.. _method-multilevel_polynomial_chaos-metric_scale:

""""""""""""
metric_scale
""""""""""""


define scaling of statistical metrics when adapting UQ surrogates


.. toctree::
   :hidden:
   :maxdepth: 1

   method-multilevel_polynomial_chaos-metric_scale-relative
   method-multilevel_polynomial_chaos-metric_scale-absolute


**Specification**

- *Alias:* None

- *Arguments:* None

- *Default:* relative


**Child Keywords:**

+-------------------------+--------------------+--------------------+-----------------------------------------------+
| Required/Optional       | Description of     | Dakota Keyword     | Dakota Keyword Description                    |
|                         | Group              |                    |                                               |
+=========================+====================+====================+===============================================+
| Required (Choose One)   | scale of           | `relative`__       | relative scaling of statistical metrics when  |
|                         | refinement metric  |                    | adapting UQ surrogates                        |
|                         |                    +--------------------+-----------------------------------------------+
|                         |                    | `absolute`__       | use of absolute statistical metrics when      |
|                         |                    |                    | adapting UQ surrogates                        |
+-------------------------+--------------------+--------------------+-----------------------------------------------+

.. __: method-multilevel_polynomial_chaos-metric_scale-relative.html
__ method-multilevel_polynomial_chaos-metric_scale-absolute.html



**Description**


When adapting a UQ surrogate model, e.g., by change in
response covariance matrix or level mappings, this option controls the
use of scaling.  The ``absolute`` option indicates that there is no
scaling such that the statistical change is compared directly to the
convergence tolerance.  The ``relative`` option indicates that the
change is measured relative to the current reference value (which
evolves during the refinement), such that the convergence tolerance
controls the relative change.

*Default Behavior*

The default is ``relative``.

*Expected Output*

*Usage Tips*

*Additional Discussion*