.. _method-multilevel_sampling-qoi_aggregation: """"""""""""""" qoi_aggregation """"""""""""""" Aggregation strategy for the QoIs statistics for problems with multiple responses in the MLMC algorithm .. toctree:: :hidden: :maxdepth: 1 method-multilevel_sampling-qoi_aggregation-sum method-multilevel_sampling-qoi_aggregation-max **Specification** - *Alias:* None - *Arguments:* None - *JSON argument key:* None **Child Keywords:** +-------------------------+------------------------+--------------------+--------------------------------+------------------------------------------------------------+ | Required/Optional | Description of Group | JSON Group Key | Dakota Keyword | Dakota Keyword Description | +=========================+========================+====================+================================+============================================================+ | Optional | | | `sum`__ | Aggregate the variances over all QoIs to generate a target | | | | | | for each level in a MLMC algorithm. | +-------------------------+------------------------+--------------------+--------------------------------+------------------------------------------------------------+ | Optional | | | `max`__ | Compute sample allocation for each response and use | | | | | | maximum over responses for each level in a MLMC algorithm | +-------------------------+------------------------+--------------------+--------------------------------+------------------------------------------------------------+ .. __: method-multilevel_sampling-qoi_aggregation-sum.html __ method-multilevel_sampling-qoi_aggregation-max.html **Description** In the multilevel method a variance of the ``allocation_target`` is computed for each of the responses and their levels :math:`Y^i_\ell, i = 1,..., R, \ell = 0,..., L` . Setting ``qoi_aggregation`` describes the rule on how to aggregate those variances over multiple response functions. Supported options are ``sum`` (default) and ``max``. For ``sum``, the variances are aggregated and a single sample allocation is computed. For ``max``, an individual sample allocation for each response using the respective variances over levels is computed and the maximum over all responses for each level is taken (worst case scenario allocation). *Default Behavior* "sum" **Examples** The following method block .. tab:: Freeform .. code-block:: dakota method, model_pointer = 'HIERARCH' multilevel_sampling pilot_samples = 20 seed = 1237 convergence_tolerance = .01 allocation_target = mean qoi_aggregation = sum .. tab:: JSON .. code-block:: json { "method": [ { "multilevel_sampling": { "model_pointer": "HIERARCH", "pilot_samples": [20], "seed_sequence": [1237], "convergence_tolerance": { "value": 0.01 }, "allocation_target": { "mean": true }, "qoi_aggregation": { "sum": true } } } ] } uses the sum rule to aggregate the variance over the qois.