Abstract:We investigate R-optimal designs for multi-response regression models with multi-factors, where the random errors in these models are correlated. Several theoretical results are derived for Roptimal designs, including scale invariance, reflection symmetry, line and plane symmetry, and dependence on the covariance matrix of the errors. All the results can be applied to linear and nonlinear models. In addition, an efficient algorithm based on an interior point method is developed for finding R-optimal designs on discrete design spaces. The algorithm is very flexible, and can be applied to any multi-response regression model.
What problem does this paper attempt to address?
The problem that this paper attempts to solve is to find R - optimal designs in multi - response regression models. Specifically, the paper focuses on the R - optimal design problem in linear and nonlinear regression models where there are correlations among multiple response variables under the influence of multiple factors. R - optimal design aims to minimize the q - dimensional rectangular volume based on the Bonferroni t - interval, which is different from the traditional D - optimal design (minimizing the volume of the confidence ellipsoid). The paper solves this problem through theoretical derivation and algorithm development, and its main contributions include:
1. **Theoretical Results**: The paper derives several theoretical properties of R - optimal designs, such as scale invariance, reflection symmetry, linear and plane symmetry, and the relationship with the error covariance matrix. These properties are applicable not only to linear models but also to nonlinear models.
2. **Algorithm Development**: The paper proposes an efficient numerical algorithm based on the interior - point method to find R - optimal designs in the discrete design space. This algorithm is very flexible and can be applied to any multi - response regression model.
3. **Application Examples**: The paper shows how to use the proposed algorithm to find R - optimal designs through several specific examples and compares the performance of this algorithm with that of the multiplicative algorithm. The results show that for complex models and larger design spaces, the proposed algorithm performs better.
### Specific Problem Description
In multi - response experimental design, multiple correlated response variables are observed in each experimental unit. For example, multiple indicators are measured from each experimental unit at a certain time point, or a single indicator is measured from each experimental unit at multiple time points. This multi - response experimental design is very common in fields such as biomedicine, pharmacy, chemistry, and engineering.
The paper studies the R - optimal design problem of multi - response regression models with multiple factors, where the random errors of the models are correlated. The model form is as follows:
\[ y_{ij} = (f_i(x_j))^{\top} \beta_i + \epsilon_{ij}, \quad i = 1, \ldots, m, \quad j = 1, \ldots, n, \]
where:
- \( y_{ij} \) is the \( i \) - th response variable on the \( j \) - th experimental unit,
- \( x_j \) is a \( p \) - vector, representing the values of \( p \) design variables on the \( j \) - th experimental unit,
- \( f_i(x) \) is a \( q_i \) - vector of known functions of \( x \),
- \( \beta_i \in \mathbb{R}^{q_i} \) is an unknown vector of regression parameters,
- \( \epsilon_{ij} \) is a random error with a mean of zero.
It is assumed that the observations of different experimental units are independent. The goal of the paper is to find the design that minimizes the R - optimal loss function, that is, to minimize the q - dimensional rectangular volume based on the Bonferroni t - interval.
### Solutions
1. **Theoretical Derivation**:
- Derived properties such as scale invariance, reflection symmetry, linear and plane symmetry of R - optimal designs.
- Analyzed the relationship between R - optimal design and the error covariance matrix, especially when the response functions are the same, the R - optimal design does not depend on the error covariance matrix.
2. **Algorithm Development**:
- Proposed an efficient numerical algorithm based on the interior - point method to find R - optimal designs in the discrete design space.
- The algorithm deals with inequality constraints by solving convex optimization problems with logarithmic barrier functions.
3. **Application Examples**:
- Demonstrated the performance of the algorithm through specific examples and compared it with the multiplicative algorithm.
- The results show that for complex models and larger design spaces, the proposed algorithm performs better.
In summary, through theoretical analysis and algorithm development, this paper systematically solves the R - optimal design problem in multi - response regression models.