Workshop on Sample Size Planning for
Intensive Longitudinal Studies

Ginette Lafit, Jordan Revol, Mihai A. Constantin, & Eva Ceulemans

DOI

📝 Description

In recent years the popularity of procedures to collect intensive longitudinal data such as the Experience Sampling Method has increased immensely. The data collected using such designs allow researchers to study the dynamics of psychological processes, and how these dynamics differ across individuals. A fundamental question when designing a study is how to determine the sample size, which is closely related to the replicability and generalizability of empirical findings. Even though multiple statistical guidelines are available for sample size planning, it still remains a demanding enterprise in complex designs. The goal of this workshop is to address this crucial question by presenting methodological advances for sample size planning for intensive longitudinal designs. First, we provide an overview of methods for sample size planning with special emphasis on a priori power analysis. Second, we focus on how to conduct power analysis in the \(N = 1\) case when the goal is to model within-person processes using \(\text{VAR}(1)\) models. Subsequently, we consider the extension to multilevel data in which multiple individuals are measured over time. We introduce an approach for conducting power analysis for multilevel models that explicitly accounts for the temporal dependencies that characterize the data collected in intensive longitudinal studies. In addition, we showcase how to perform power analysis for these models using a user-friendly and open-source application. Finally, we consider an alternative criterion for conducting sample size planning that targets the predictive accuracy of a model for unseen data. Focusing on \(\text{VAR}(1)\) models in an \(N = 1\) context, we introduce a novel approach, called predictive accuracy analysis, to assess how many measurement occasions are required in order to optimize predictive accuracy.

📊 Learning Objectives

The workshop provides a road map on how to determine the sample size in intensive longitudinal designs. Upon course completion, participants will:

  • be familiar with methods for conducting power analysis for \(\text{AR}(1)\) and \(\text{VAR}(1)\) models in \(N = 1\) and multilevel intensive longitudinal designs
  • understand the key differences between simulation-based and analytical power analysis approaches
  • be able to leverage existing tools for conducting power analysis for \(\text{AR}(1)\) and \(\text{VAR}(1)\) for intensive longitudinal designs
  • be familiar with new methods for conducting sample size analysis based on criteria different than statistical power (e.g., predictive accuracy or sensitivity)

💻 Prerequisites

Participants should have some basic knowledge of R and some experience with RStudio. For the hands-on parts of the workshop, you need to install R version 4.1.2 or higher, RStudio, and several R packages as indicated on the page corresponding to each exercise.

Some exercises in this workshop also involve using Shiny applications to run power analysis. You can find additional instructions on how to download and run the Shiny applications below:

You can find detailed instructions and examples for conducting sample size analysis using the powerly package at powerly.dev.

📂 Modules

Topic Duration Slides Tutorial
Introduction to sample size planning in intensive longitudinal research 45m slides -
Sample size planning for \(\text{VAR}(1)\) models in \(N = 1\) designs 60m slides tutorial 1 tutorial 2
Sample size planning for multilevel models applied to intensive longitudinal designs 50m slides tutorial 1 tutorial 2 tutorial 3
Advanced methods for sample size analysis 40m slides tutorial

📍 Given At

Conference Location Date Link
SAA 2023 Amsterdam, The Netherlands June 8th, 2023 link

✍️ Citation

⚖️ License