Longitudinal Study Strategies

Had to share the one prompt that has transformed how I approach AI research. 📌 Save this post. Don’t just ask for point-in-time data like a junior PM. Instead, build in more temporal context through systematic data collection over time. Use this prompt to become a superforecaster with the help of AI. Great for product ideation, competitive research, finance, investing, etc. ⏰⏰⏰⏰⏰⏰⏰⏰⏰⏰⏰⏰ TIME MACHINE PROMPT: Execute longitudinal analysis on [TOPIC]. First, establish baseline parameters: define the standard refresh interval for this domain based on market dynamics (enterprise adoption cycles, regulatory changes, technology maturity curves). For example, AI refresh cycle may be two weeks, clothing may be 3 months, construction may be 2 years. Calculate n=3 data points spanning 2 full cycles. For each time period, collect: (1) quantitative metrics (adoption rates, market share, pricing models), (2) qualitative factors (user sentiment, competitive positioning, external catalysts), (3) ecosystem dependencies (infrastructure requirements, complementary products, capital climate, regulatory environment). Structure output as: Current State Analysis → T-1 Comparative Analysis → T-2 Historical Baseline → Delta Analysis with statistical significance → Trajectory Modeling with confidence intervals across each prediction. Include data sources. ⏰⏰⏰⏰⏰⏰⏰⏰⏰⏰⏰⏰

How to capture subject-specific temporal trajectories from longitudinal multivariate data reliably? Longitudinal datasets are collected in many fields with the goal of extracting insights, capturing early risk markers of diseases or understanding various conditions, e.g., exposures in early life. For instance, blood metabolites are measured over time during meal challenge tests to understand differences in metabolic responses of individuals and how those are related to cardiometabolic diseases. Similarly, sensitization to allergens during childhood has been studied to investigate associations with atopic diseases. Despite differences in data characteristics, there is the common goal of capturing and understanding individual differences. For instance, sensitization to certain food allergens has been shown to increase in early childhood and then decrease. The temporal trajectory, however, is not necessarily the same for everyone, and differences may reveal important insights. Therefore, there is an emerging need for data science methods that can extract subject-specific temporal patterns from longitudinal multivariate data reliably. We show that coupled matrix factorizations (CMF) are effective tools to capture such subject-specific temporal patterns. In collaboration with COPSAC (COPSAC2000 cohort), we focus on two novel applications: analysis of longitudinal metabolomics data and sensitization data. In metabolomics, by accounting for subject variability in temporal profiles, CMF models reveal differences in the metabolic response of individuals (in a postprandial meal challenge) according to anthropometric and insulin sensitivity measures. In sensitization data analysis, CMF models reveal differences in temporal trajectories of children according to delivery/birth mode, i.e., children born by C-section show earlier sensitization to a specific group of food allergens compared to those born by natural birth. We discuss the reliability of extracted patterns in terms of reproducibility and replicability. Christos Chatzis, David Horner, Rasmus Bro, Ann-Marie Malby Schoos, Morten Arendt Rasmussen, Evrim Acar Ataman, Revealing Subject-Specific Temporal Patterns from Longitudinal Data, bioRxiv, https://lnkd.in/efgCF_Vz

There are two common methods to test hypotheses about treatment effect in 𝗹𝗼𝗻𝗴𝗶𝘁𝘂𝗱𝗶𝗻𝗮𝗹 studies in randomized clinical trials (RCTs): 𝗠𝗠𝗥𝗠 and 𝗰𝗟𝗗𝗔. 𝗠𝗠𝗥𝗠 (Mixed-Model for Repeated Measurements), despite its name, contains only fixed effects and is a marginal model (unlike true mixed models, giving outcomes conditional to random effects, e.g. patients). Typically it's fitted using Generalized Least Square (Gaussian response) or Generalized Estimating Equations (Gaussian but problematic residuals, non-Gaussian GLM). For RCTs we assume either of two models (equivalent) in R notation: POST_measur ~ 1+ TRT+ Time + TRT : Time + PRE_measur + PRE_measur : Time. CHANGE_measur ~ 1+ TRT+ Time + TRT : Time + PRE_measur + PRE_measur : Time. 𝗰𝗟𝗗𝗔 (constrained Longitudinal Data Analysis) is a special kind of MMRM or a mixed model where the baseline means across both treatment arms are constrained to be the same. It has a "weird-looking" formula, without intercept and baseline treatment variable. It's quite common in clinical trials and is (usually) more powerful than MMRM or (single-point) ANCOVA (analysis of post-treatment response adjusted for its pre-treatment value). It may show elevated type-1 error. It was proposed in ~2000 by Liang and Zeger (yes, you remember well - invented the GEE estimation). Any_time_measur ~ 0 + Time + Treatment : Time But the selection of analytical method can affect the analysis principle. I assume you are familiar with the ITT (Intention-To-Treat principle). Let's think how the two models differ in this context: 👉 𝗠𝗠𝗥𝗠 puts the POST-treatment measurements (or change from from baseline, CFB, gain score, etc.) into the response variable (Y) and adjusts for the pre-treatment (baseline). What does it mean effectively? 💡 That patient must have at least one follow-up (POST-intervention) assessment to enter the analysis. It's not the FAS (Full Analysis Set). This is the 𝗺𝗼𝗱𝗶𝗳𝗶𝗲𝗱 𝗜𝗻𝘁𝗲𝗻𝘁𝗶𝗼𝗻-𝘁𝗼-𝗧𝗿𝗲𝗮𝘁 (𝗺𝗜𝗧𝗧) approach: randomized + screening assessment + at least one post-intervention assessment. --- 👉 𝗰𝗟𝗗𝗔 puts the all measurements (including the baseline) into the response variable (Y) and employ appropriate constraint. What does it mean effectively? 💡 That patient does not have anything but screening measurement and randomization. This is the 𝗜𝗻𝘁𝗲𝗻𝘁𝗶𝗼𝗻-𝘁𝗼-𝗧𝗿𝗲𝗮𝘁 (𝗜𝗧𝗧) approach: randomized + screening assessment. If the screening is missing, it should be at least MAR. Take-home message: 🎯 If you use a mixed model or MMRM applied to post-treatment variables in RCT, you effectively do the 𝗺𝗜𝗧𝗧 analysis. 🎯 If you want the 𝗜𝗧𝗧 analysis, try the cLDA (or the ordinary LDA with baseline included in the response, but this disables the baseline adjustment, which is listed by the regulatory guidelines). #clinicaltrials #clinicalresearch #biostatistics #statistics

𝗪𝗵𝗮𝘁 𝗰𝗮𝗻 𝘄𝗲 𝗹𝗲𝗮𝗿𝗻 𝗳𝗿𝗼𝗺 𝗿𝗲𝘀𝗲𝗮𝗿𝗰𝗵𝗲𝗿𝘀 𝗮𝗯𝗼𝘂𝘁 𝗲𝘃𝗮𝗹𝘂𝗮𝘁𝗶𝗻𝗴 𝗼𝘂𝗿 𝗽𝗿𝗮𝗰𝘁𝗶𝗰𝗲? Researchers don't just trust their intuition about whether their work is effective. They systematically gather data, track patterns over time, and use multiple sources of evidence to understand what's really happening. As practitioners - whether we're coaches, consultants, trainers, therapists, or educators - we can borrow this disciplined and systematic approach to move beyond post-session feelings and anecdotal success stories. By treating our practice as worthy of careful examination, we can gather longitudinal data through client or participant feedback, track measurable indicators of progress relevant to our field, and analyze patterns that reveal what's actually working. This doesn't mean turning practice into a clinical trial. It means being intentional about gathering evidence that helps us see our work clearly - making visible what's often invisible in our effectiveness. 𝗛𝗲𝗿𝗲'𝘀 𝗵𝗼𝘄 𝘁𝗼 𝘀𝘁𝗮𝗿𝘁: First, define what "effective" means in your context. Without clear criteria, you're measuring fog. What are 2-3 concrete indicators of success in your practice? Second, establish a baseline at the start of each engagement. You can't measure change without knowing the starting point. Third, create brief feedback mechanisms - simple forms or check-ins that take 5-10 minutes maximum. Regular data collection reveals patterns over time, but only if people actually give you data. Google Forms, or something similar, can help with this. Fourth, build feedback collection into your process from the beginning. Make it a normal part of how you work, not an awkward add-on. Fifth, analyze for patterns. Review your collected data regularly, looking for trends across clients or participants. Tools like Dedoose can help organize and analyze both numerical ratings and open-ended responses, making it easier to spot what's working and where your practice has blind spots. Finally, act on what you learn. Data without action is just interesting. Use these insights to adjust your approach, seek development in areas of weakness, or double down on what's working. This systematic approach doesn't replace professional intuition or the art of practice. It enhances it by giving you evidence to support, challenge, or refine what your experience tells you. #PracticeEvaluation #EvidenceBasedPractice #QualitativeResearch #ProfessionalDevelopment #ICFCoach

Working with clinical researchers, I often see the same challenge come up: how to properly analyze longitudinal data. A longitudinal study simply means that the same group of subjects is measured repeatedly over time. For example, in a cosmetic study, you might follow two cohorts (treatment vs control) and measure skin hydration at several time points. One question you may ask: At which time point does treatment induce a significantly higher skin hydration compared to the control? In practice, many analyses still rely on: - Independent t-tests (or Mann-Whitney's test): comparing cohorts at a given time point. - Paired t-tests (or Wilcoxon's test): comparing the same cohort between two time points. While useful, these approaches only capture part of the story. 👉 A more robust and efficient alternative exists: mixed-effects models. Why use them? - Mixed models account for the correlation between repeated measurements within the same subject. - They allow you to model time and group effects simultaneously. - They handle missing data much more gracefully. The good news: these models are now easily accessible in JASP, an open-source software developed at the University of Amsterdam. 🎥 I’ve put together a short demo showing how to run a mixed model in JASP. Do not hesitate to pause and check captions ;) Have you already used JASP in medical research? Tell me in the comments 👇. ---- Mixed models are powerful—but they often require some guidance to get started properly. If you need support with statistical analyses for your research projects, feel free to reach out. #statistics #clinicalresearch #biostatistics #datascience

What happens when you have a wonderful research team (William E. Donald, Helen Hughes, Rebecca Padgett, SFHEA and Maria Mouratidou), you all work hard on a project that you hope will get into one of the top-ranked journals, yet the dataset is somewhat disappointing? First, publish the empirical findings anyway because they still have practitioner value (thanks, Higher Education, Skills and Work-Based Learning, for giving us a home for that paper). Second, write another paper setting out some of the challenges faced and how other researchers might overcome them. Just published online in Management Research Review, we present... Balancing rigour and practicality when conducting large-scale, longitudinal audio diary studies: Guidance for business and management research Purpose This paper aims to expand the methodological toolkit by detailing an approach for conducting large-scale audio diary studies in business and management education research. The authors address two research questions: (1) How can large-scale, longitudinal audio diary studies be conducted in a way that balances rigour and practicality? (2) What methodological challenges arise when implementing a longitudinal audio diary study at scale, and how might these be addressed? Design/methodology/approach They draw on insights from a study with 128 international students in a UK business school who provided a combined 602 reflective audio diary contributions to inform a ten-step process for conducting large-scale audio diary studies. Findings They present a flexible ten-step process, discussing challenges and recommendations regarding participant engagement and retention, technology, data quality and resource demands. Originality/value They conclude with examples of four themes for future research applications of audio diary studies: employability perceptions, mental health and well-being, ethical decision-making and corporate social responsibility awareness and innovation and entrepreneurial mindset development. These research avenues represent opportunities to promote a sustainable career trajectory for individuals and offer societal benefits via a sustainable career ecosystem. To Cite: Donald, W. E., Hughes, H. P. N., Padgett, R. C., & Mouratidou, M. (2025). Balancing rigour and practicality when conducting large-scale, longitudinal audio diary studies: Guidance for business and management research. Management Research Review. Advanced Online Publication. DOI: 10.1108/MRR-03-2025-0210. Links to the publisher version (behind a paywall) and the AAM (free version as accepted without formatting to journal style) in the comments. P.S. No idea why they changed "We" to "The Authors" and subsequently "They" during production, as that was not the case in the proofs. Oh well!