FAQ

Why doesn’t mUQSA support Quantity of Interest (QoI) with dimensions greater than 2? What if I have a QoI with more than 2 dimensions?
Get answermUQSA primarily focuses on univariate and bivariate sensitivity analysis. Handling multi-dimensional QoIs can be complex and resource-intensive. In addition, it's important to note that visualizing sensitivity results in more than two dimensions can be ineffective and challenging. If your Quantity of Interest (QoI) has more than two dimensions, you may consider transforming your data into two-dimensional form. If your analysis requires assessing sensitivity in higher-dimensional spaces, you may need to develop custom solutions or use specialized software designed for higher-dimensional sensitivity analysis.
How can I ensure that the results generated by mUQSA are reliable?
Get answerTo ensure the reliability of results generated by mUQSA, consider the following best practices:

Verification and Validation: Ensure that your models are verified and validated for accuracy and correctness.

Sensitivity Analysis Settings: Carefully configure mUQSA’s settings, including the number of samples and sensitivity analysis method.

Model Uncertainty: Account for uncertainties in your models and input parameters.

Convergence Checks: Verify that sensitivity indices have converged by increasing the number of samples if necessary.

Interpretation: Understand the limitations of each sensitivity analysis method and interpret results in the context of your specific problem.

By following these steps, you can improve the reliability of the sensitivity analysis results produced by mUQSA.

Why does mUQSA offer multiple sensitivity analysis methods?
Get answermUQSA provides multiple sensitivity analysis methods to cater to different user needs and problem types. Each method has its strengths and limitations. Users can choose the most appropriate method based on their specific goals, the complexity of the model, and the available computational resources. The diversity of methods in mUQSA allows users to perform sensitivity analysis in a way that best suits their requirements, making it a versatile tool for a wide range of applications in uncertainty quantification and sensitivity analysis.

When conducting sensitivity analysis with mUQSA, it’s worth considering the Monte Carlo (MC) method as a suitable starting point. MC is less sensitive to the number of parameters, making it a versatile choice for a wide range of scenarios. This method can be used effectively in sensitivity analysis to discover which parameters are less important, and therefore, which ones could potentially be omitted from further analysis. For a more detailed analysis of a selected few parameters, you may consider usage of PCE or SC.

How can I deal with the curse of dimensionality in sensitivity analysis when I have a large number of input parameters?
Get answerThe curse of dimensionality, which refers to the challenges of dealing with high-dimensional input spaces, can be addressed by using dimension reduction techniques or surrogate models. Dimension reduction techniques like Principal Component Analysis (PCA) or Partial Least Squares (PLS) can help you identify and focus on the most influential parameters. Unfortunately, right now, mUQSA does not support dimension reduction techniques nor surrogate models. If you want you can apply these mechanisms directly in your model.
Is mUQSA suitable for sensitivity analysis of models with proprietary code?
Get answerYes, mUQSA implements so-called non-intrusive UQ and treats models as black-boxes, so it can be used for sensitivity analysis of proprietary code. It does not require access to the model's internal workings but relies on input-output data. You can use mUQSA to analyze the model's behavior without needing to know its source code or internal structure.
Can mUQSA handle sensitivity analysis for models that involve time-consuming and resource-consuming simulations?
Get answerOne of the significant advantages of mUQSA is its seamless integration with supercomputer environment. This integration allows mUQSA to efficiently handle sensitivity analysis for large-scale models. Consequently, computational resources of supercomputer can easily be employed to enhance the speed and capability of sensitivity analysis performed with the platform. However, the efficiency and practicality of the analysis still depend on the specific sensitivity analysis method chosen and its configuration.