Polynomial Chaos Expansion for general multivariate distributions with correlated variables

Maria Navarro,Jeroen Witteveen,Joke Blom

DOI: https://doi.org/10.48550/arXiv.1406.5483

2014-06-21

Abstract:Recently, the use of Polynomial Chaos Expansion (PCE) has been increasing to study the uncertainty in mathematical models for a wide range of applications and several extensions of the original PCE technique have been developed to deal with some of its limitations. But as of to date PCE methods still have the restriction that the random variables have to be statistically independent. This paper presents a method to construct a basis of the probability space of orthogonal polynomials for general multivariate distributions with correlations between the random input variables. We show that, as for the current PCE methods, the statistics like mean, variance and Sobol' indices can be obtained at no significant extra postprocessing costs. We study the behavior of the proposed method for a range of correlation coefficients for an ODE with model parameters that follow a bivariate normal distribution. In all cases the convergence rate of the proposed method is analogous to that for the independent case. Finally, we show, for a canonical enzymatic reaction, how to propagate experimental errors through the process of fitting parameters to a probabilistic distribution of the quantities of interest, and we demonstrate the significant difference in the results assuming independence or full correlation compared to taking into account the true correlation.

Numerical Analysis

What problem does this paper attempt to address?

The problem that this paper attempts to solve is the problem of dealing with multivariate random inputs with correlations in the Polynomial Chaos Expansion (PCE) method. The traditional PCE method requires statistical independence among random input variables, which limits its flexibility and accuracy in practical applications. This paper proposes a method for constructing an orthogonal polynomial basis in the probability space applicable to general multivariate distributions (including random input variables with correlations), thus overcoming this limitation.

Specifically, the main contributions of the paper include:

1. **Constructing an orthogonal polynomial basis applicable to correlated random inputs**: The paper proposes a method to construct an orthogonal polynomial basis applicable to multivariate distributions with correlations. This method is based on the Gram - Schmidt orthogonalization method and can construct a set of orthogonal polynomial bases starting from any set of linearly independent polynomials.

2. **Efficient calculation of statistics**: The paper shows how to efficiently calculate statistics such as the mean, variance, and Sobol' index in this new PCE method. The calculation cost of these statistics is comparable to that of the traditional PCE method and does not significantly increase the additional post - processing cost.

3. **Verifying the effectiveness of the method**: Through an example of an ordinary differential equation (ODE), the paper verifies the effectiveness of the proposed method. The behavior of the solution under different correlation coefficients is studied, and it is shown that the convergence rate is the same as that in the independent case.

4. **Application examples**: The paper also shows, through a classical enzymatic reaction model, how to propagate experimental errors through parameter fitting to the probability distribution of the quantity of interest (such as product concentration). The results show that considering the real correlation has a significant impact on the final result, and there are obvious differences compared with the cases of assuming independence or complete correlation.

In summary, this paper aims to extend the PCE method so that it can handle multivariate random inputs with correlations, thereby improving the accuracy and application range of uncertainty quantification analysis.