Abstract:Hypothesis testing for the slope function in functional linear regression is of both practical and theoretical interest. We develop a novel test for the nullity of the slope function, where testing the slope function is transformed into testing a high-dimensional vector based on functional principal component analysis. This transformation fully circumvents ill-posedness in functional linear regression, thereby enhancing numeric stability. The proposed method leverages the technique of bootstrapping max statistics and exploits the inherent variance decay property of functional data, improving the empirical power of tests especially when the sample size is limited or the signal is relatively weak. We establish validity and consistency of our proposed test when the functional principal components are derived from data. Moreover, we show that the test maintains its asymptotic validity and consistency, even when including \emph{all} empirical functional principal components in our test statistics. This sharply contrasts with the task of estimating the slope function, which requires a delicate choice of the number (at most in the order of $\sqrt n$) of functional principal components to ensure estimation consistency. This distinction highlights an interesting difference between estimation and statistical inference regarding the slope function in functional linear regression. To the best of our knowledge, the proposed test is the first of its kind to utilize all empirical functional principal components.
What problem does this paper attempt to address?
The paper attempts to address the problem of hypothesis testing for the slope function in Functional Linear Regression (FLM). Specifically, the paper focuses on how to test whether the slope function is zero, i.e., testing the following hypothesis:
\[ H_0: \beta = 0 \quad \text{vs.} \quad H_a: \beta \neq 0 \]
### Background and Motivation
In Functional Data Analysis (FDA), the Functional Linear Model (FLM) is an important tool for studying the linear relationship between a response variable and a predictor variable, where at least one of the variables is in functional form. These models are very common in many practical applications, such as medicine, meteorology, and finance.
### Limitations of Existing Methods
Existing methods face several challenges when dealing with this problem:
1. **Ill-posedness**: Directly estimating the coefficients of the slope function usually leads to ill-posedness because these coefficients involve the reciprocals of eigenvalues, which tend to approach zero rapidly in functional data.
2. **Testing power with limited sample size or weak signals**: Existing methods may have low testing power when the sample size is limited or the signal is weak.
### Main Contributions of the Paper
To overcome the above challenges, the paper proposes a new testing method with the following features:
1. **Transformation to high-dimensional vector testing**: Through Functional Principal Component Analysis (FPCA), the problem of testing the slope function is transformed into a problem of testing high-dimensional vectors. This method completely avoids the ill-posedness issue and improves numerical stability.
2. **Utilization of all empirical principal components**: The new method can utilize all empirical principal components, without the need to select a specific number of principal components as required by existing methods. This not only simplifies the tuning process but also potentially improves the testing power, especially when the signal of the slope function is related to higher-order principal components.
3. **Bootstrap max statistics**: By using bootstrapping to construct max statistics and leveraging the inherent variance decay property of functional data, the method improves empirical testing power, particularly in cases with limited sample size or weak signals.
### Theoretical Results
The paper also establishes the validity and consistency of the proposed testing method and proves that even when using all empirical principal components, the testing method still maintains its asymptotic validity and consistency. This finding highlights an important distinction between estimation and statistical inference in functional linear regression.
### Experimental Validation
The paper demonstrates the numerical performance of the proposed method through simulation studies and real-world applications, validating its effectiveness and superiority in practical applications.
In summary, the paper proposes a novel and effective hypothesis testing method that addresses the key issue of testing the slope function in functional linear regression, providing new tools and insights for research in related fields.