Optimal Designs for Longitudinal and Functional Data

TL;DR

This paper develops optimal measurement designs for longitudinal and functional data collection, improving prediction accuracy from limited data by strategically selecting measurement times based on prior pilot studies.

Contribution

It introduces novel optimal design methods for longitudinal and functional data, leveraging prior pilot study information to enhance data collection efficiency.

Findings

Optimal designs significantly outperform random measurement points.

Designs achieve asymptotic optimality.

Simulations and real data demonstrate improved prediction accuracy.

Abstract

We propose novel optimal designs for longitudinal data for the common situation where the resources for longitudinal data collection are limited, by determining the optimal locations in time where measurements should be taken. As for all optimal designs, some prior information is needed to implement the proposed optimal designs. We demonstrate that this prior information may come from a pilot longitudinal study that has irregularly measured and noisy measurements, where for each subject one has available a small random number of repeated measurements that are randomly located on the domain. A second possibility of interest is that a pilot study consists of densely measured functional data and one intends to take only a few measurements at strategically placed locations in the domain for the future collection of similar data. We construct optimal designs by targeting two criteria: (a) Optimal designs to recover the unknown underlying smooth random trajectory for each subject from a few optimally placed measurements such that squared prediction errors are minimized; (b) Optimal designs that minimize prediction errors for functional linear regression with functional or longitudinal predictors and scalar responses, again from a few optimally placed measurements. The proposed optimal designs address the need for sparse data collection when planning longitudinal studies, by taking advantage of the close connections between longitudinal and functional data analysis. We demonstrate in simulations that the proposed designs perform considerably better than randomly chosen design points and include a motivating data example from the Baltimore longitudinal study of aging. The proposed designs are shown to have an asymptotic optimality property.