Is There a Secret Recipe Between Curves? Let's Find Out How Statistics Discovers It

Imagine you have two graphs, perhaps one showing how a plant grows and another how temperature changes. They seem to have nothing in common, but maybe there's a hidden link, like a 'secret recipe' that transforms one graph into the other. Scientists have created a statistical test specifically to find this arbitrary transformation. The method is based on estimating the second derivative of this unknown transformation, analyzing the data with advanced techniques from empirical process theory. To work even with small datasets, they developed a bootstrap algorithm that repeats the test many times to check the reliability of the result. They tested the method on simulated data and on real brain tissue imaging (DTI) data, showing that it can reveal non-obvious relationships between curves.

From a technical standpoint, the core of the work is the formulation of a null hypothesis test asserting the absence of an arbitrary transformation between the slope functions of two scalar-on-function regression models. The test statistic is built starting from the estimate of the second derivative of the unknown transformation, a choice that allows capturing the local curvature of the relationship. The asymptotic properties of the statistic are studied using empirical process theory, ensuring that, with sufficiently large samples, the distribution of the statistic is known and predictable. For practical applications with small sample sizes, the bootstrap algorithm is fundamental: it allows simulating the distribution of the statistic under the null hypothesis, providing a reliable p-value without depending on the asymptotic approximation. The application to DTI (Diffusion Tensor Imaging) data is particularly relevant, as it allows analyzing how microstructural properties of brain white matter can be linked across different regions through non-linear and non-prescribed relationships.