Sensitivity analysis methods in the biomedical sciences

George Qian,Adam Mahdi

DOI: https://doi.org/10.48550/arXiv.2001.03965

2020-01-13

Abstract:Sensitivity analysis is an important part of a mathematical modeller's toolbox for model analysis. In this review paper, we describe the most frequently used sensitivity techniques, discussing their advantages and limitations, before applying each method to a simple model. Also included is a summary of current software packages, as well as a modeller's guide for carrying out sensitivity analyses. Finally, we apply the popular Morris and Sobol methods to two models with biomedical applications, with the intention of providing a deeper understanding behind both the principles of these methods and the presentation of their results.

Quantitative Methods,Applications

What problem does this paper attempt to address?

The main problem that this paper attempts to solve is: how to effectively conduct sensitivity analysis (SA) in biomedical science to understand and quantify the impact of model input uncertainty on the output. Specifically, the author aims to:

1. **Introduce and evaluate commonly - used sensitivity analysis methods**: The paper describes in detail several of the most commonly - used sensitivity analysis techniques, discusses their advantages and limitations, and shows the specific application process of these methods through the application of simple models.

2. **Provide practical guidelines for sensitivity analysis**: In order to help researchers better apply sensitivity analysis, the paper not only summarizes the currently available relevant software packages, but also provides a detailed step - by - step guide on how to conduct sensitivity analysis in actual research.

3. **Gain an in - depth understanding of the principles of sensitivity analysis and the presentation of its results**: Through the application of the Morris method and the Sobol method in two models with biomedical applications, the paper hopes that readers can gain a more in - depth understanding of how these methods work and how to interpret their results.

4. **Address the challenges in biomedical modeling**: Models in the biomedical field usually have a large number of free parameters, and the values of these parameters may affect the behavior and interpretation of the model. Sensitivity analysis can help ensure the identifiability of the model, thereby achieving accurate and meaningful parameter inference. In addition, the inherent randomness of biological processes and the uncertainty of data also make sensitivity analysis particularly important.

### Mathematical expression of specific problems

The core problem of sensitivity analysis is to quantify how the uncertainty of model output is allocated to input uncertainties from different sources. Mathematically, sensitivity analysis can be defined as:

\[
S_i=\frac{\partial Y}{\partial X_i}\bigg|_{X = x^*}
\]

where \(Y\) is the model output, \(X_i\) is the \(i\) - th input factor, and \(x^*\) is the evaluation point. For global sensitivity analysis, especially variance - based methods such as the Sobol method, the model variance will be further decomposed:

\[
S_i=\frac{V_{X_i}(E_{X\sim i}(Y|X_i))}{V(Y)}
\]

Here, \(E_{X\sim i}(Y|X_i)\) represents the expected value of all other input factors \(X\sim i\) under the condition of a given \(X_i\), and \(V_{X_i}\) represents the variance with respect to \(X_i\).

Through these methods, the paper hopes to provide a systematic sensitivity analysis framework for biomedical science research, helping researchers better understand and optimize their models.