4.4 Article

ANCOVA Versus CHANGE From Baseline in Nonrandomized Studies: The Difference

Journal

MULTIVARIATE BEHAVIORAL RESEARCH
Volume 48, Issue 6, Pages 895-922

Publisher

ROUTLEDGE JOURNALS, TAYLOR & FRANCIS LTD
DOI: 10.1080/00273171.2013.831743

Keywords

-

Ask authors/readers for more resources

The pretest-posttest control group design can be analyzed with the posttest as dependent variable and the pretest as covariate (ANCOVA) or with the difference between posttest and pretest as dependent variable (CHANGE). These 2 methods can give contradictory results if groups differ at pretest, a phenomenon that is known as Lord's paradox. Literature claims that ANCOVA is preferable if treatment assignment is based on randomization or on the pretest and questionable for preexisting groups. Some literature suggests that Lord's paradox has to do with measurement error in the pretest. This article shows two new things: First, the claims are confirmed by proving the mathematical equivalence of ANCOVA to a repeated measures model without group effect at pretest. Second, correction for measurement error in the pretest is shown to lead back to ANCOVA or to CHANGE, depending on the assumed absence or presence of a true group difference at pretest. These two new theoretical results are illustrated with multilevel (mixed) regression and structural equation modeling of data from two studies.

Authors

I am an author on this paper
Click your name to claim this paper and add it to your profile.

Reviews

Primary Rating

4.4
Not enough ratings

Secondary Ratings

Novelty
-
Significance
-
Scientific rigor
-
Rate this paper

Recommended

Article Rehabilitation

Aquatic Cycling Improves Knee Pain and Physical Functioning in Patients With Knee Osteoarthritis: A Randomized Controlled Trial

Stefanie Rewald, A. F. Ton Lenssen, Pieter J. Emans, Rob A. de Bie, Gerard van Breukelen, Ilse Mesters

ARCHIVES OF PHYSICAL MEDICINE AND REHABILITATION (2020)

Letter Health Care Sciences & Services

Letter to the editor: Sufficient statistics and insufficient explanations

Gerard J. P. van Breukelen

STATISTICAL METHODS IN MEDICAL RESEARCH (2020)

Article Health Care Sciences & Services

Optimal two-stage sampling for mean estimation in multilevel populations when cluster size is informative

Francesco Innocenti, Math J. J. M. Candel, Frans E. S. Tan, Gerard J. P. van Breukelen

Summary: The article discusses estimating the mean of a quantitative variable in a hierarchical population and compares three competing two-stage sampling designs. Sampling clusters with probability proportional to size is recommended as the optimal approach.

STATISTICAL METHODS IN MEDICAL RESEARCH (2021)

Article Public, Environmental & Occupational Health

A randomized controlled trial comparing community lifestyle interventions to improve adherence to diet and physical activity recommendations: the VitalUM study

Hilde Marijke van Keulen, Gerard van Breukelen, Hein de Vries, Johannes Brug, Ilse Mesters

Summary: This study explores the effectiveness of social-cognitive interventions on improving adherence to physical activity and fruit and vegetable consumption guidelines. Self-efficacy expectations, habit strength, and stages of change were found to predict adherence behaviors, while factors like sex, awareness, and action plans also influenced adherence. The study suggests that less resource-intensive interventions may have a significant public health impact if widely implemented.

EUROPEAN JOURNAL OF EPIDEMIOLOGY (2021)

Article Computer Science, Interdisciplinary Applications

Robustness of cost-effectiveness analyses of cluster randomized trials assuming bivariate normality against skewed cost data

Md Abu Manju, Math J. J. M. Candel, Gerard J. P. van Breukelen

Summary: The bivariate normal multilevel model is widely used in cost-effectiveness analyses and cluster randomized trials, and is robust against non-normal cost distributions. Studies show that CEAs and sample size calculations based on this model are quite robust against highly skewed cost data, especially in terms of type I error rate and power.

COMPUTATIONAL STATISTICS & DATA ANALYSIS (2021)

Article Statistics & Probability

REMAXINT: a two-mode clustering-based method for statistical inference on two-way interaction

Zaheer Ahmed, Alberto Cassese, Gerard van Breukelen, Jan Schepers

Summary: REMAXINT is a novel method for capturing two-way interactions in two-mode data, estimating parameters through maximizing conditional classification likelihood, and proposing a max - F test for testing interaction effects. Through simulation studies and empirical case studies, we demonstrate the effectiveness and utility of the method.

ADVANCES IN DATA ANALYSIS AND CLASSIFICATION (2021)

Article Mathematical & Computational Biology

Maximin design of cluster randomized trials with heterogeneous costs and variances

Gerard J. P. Breukelen, Math J. J. M. Candel

Summary: Cluster randomized trials evaluate treatment effects on nested individuals within clusters, considering unknown variances and differences between treatment arms. This paper proposes a Maximin design that maximizes the minimum relative efficiency in the variance parameter space, compared with other designs.

BIOMETRICAL JOURNAL (2021)

Article Psychology, Multidisciplinary

Sample Size Calculation and Optimal Design for Regression-Based Norming of Tests and Questionnaires

Francesco Innocenti, Frans E. S. Tan, Math J. J. M. Candel, Gerard J. P. van Breukelen

Summary: To prevent errors in psychological assessment, it is important to have precise test norms. This study proposes a method for sample size planning based on regression-based norming to improve the precision of inference on Z-scores and percentile rank scores. The study derives sampling variance formulas and uses them to determine the optimal design for the normative sample. It also provides formulas to compute the normative sample size based on desired power and precision.

PSYCHOLOGICAL METHODS (2023)

Article Pharmacology & Pharmacy

Imputation of Missing Covariates in Randomized Controlled Trials with Continuous Outcomes: Simple, Unbiased and Efficient Methods

Mutamba T. Kayembe, Shahab Jolani, Frans E. S. Tan, Gerard J. P. van Breukelen

Summary: The literature suggests using sophisticated methods (such as multiple imputation and maximum likelihood) for dealing with missing covariates in nonrandomized studies, but these methods may not be optimal in randomized studies. This study extends the analysis to multiple baseline covariates with missingness and compares the performance of multiple imputation and maximum likelihood with simple alternative methods in various scenarios. The results show that all simple methods provide unbiased treatment effect estimation, but with increased mean squared residual. Mean imputation and the missing-indicator method perform best in all covariate missingness scenarios.

JOURNAL OF BIOPHARMACEUTICAL STATISTICS (2022)

Article Computer Science, Interdisciplinary Applications

Model fit criteria curve behaviour in class enumeration - a diagnostic tool for model (mis)specification in longitudinal mixture modelling

Gavin van der Nest, Valeria Lima Passos, Math J. J. M. Candel, Gerard J. P. van Breukelen

Summary: This study examines the behavior of fit criteria curves of longitudinal finite mixture models under different data conditions and covariance structures. The variability in fit statistic patterns is attributed to the level of class separation and the presence/absence of random effects. The findings support some widely held notions while debunking others, providing guidelines on using fit criteria curves as a diagnostic aid during class enumeration.

JOURNAL OF STATISTICAL COMPUTATION AND SIMULATION (2022)

Article Medicine, Research & Experimental

Laparoscopic versus open resections in the posterosuperior liver segments within an enhanced recovery programme (ORANGE Segments): study protocol for a multicentre randomised controlled trial

Christoph Kuemmerli, Robert S. Fichtinger, Alma Moekotte, Luca A. Aldrighetti, Somaiah Aroori, Marc G. H. Besselink, Mathieu D'Hondt, Rafael Diaz-Nieto, Bjorn Edwin, Mikhail Efanov, Giuseppe M. Ettorre, Krishna Menon, Aali J. Sheen, Zahir Soonawalla, Robert Sutcliffe, Roberto Troisi, Steven A. White, Lloyd Brandts, Gerard J. P. van Breukelen, Jasper Sijberden, Sian A. Pugh, Zina Eminton, John N. Primrose, Ronald van Dam, Mohammed Abu Hilal

Summary: The ORANGE Segments trial is the first multicentre international randomised controlled study to compare short- and long-term surgical and oncological outcomes of laparoscopic and open resections in the posterosuperior segments within an enhanced recovery programme.

TRIALS (2022)

Editorial Material Nutrition & Dietetics

Best (but oft forgotten) practices: Efficient sample sizes for commonly used trial designs

Math J. J. M. Candel, Gerard J. P. van Breukelen

Summary: Designing studies with high statistical power is crucial for improving the quality and reproducibility of research findings. This article presents designs that minimize the number of subjects or research budget while maintaining the desired power level. It focuses on allocation of subjects to treatments and optimal number of centers in cluster-randomized and multicenter trials. Maximin designs are introduced to address unknown parameters, ensuring a pre-specified power level and minimizing research costs.

AMERICAN JOURNAL OF CLINICAL NUTRITION (2023)

Article Mathematics, Interdisciplinary Applications

Cluster Randomized Trials with a Pretest and Posttest: Equivalence of Three-, Two- and One-Level Analyses, and Sample Size Calculation

Gerard J. P. Van Breukelen

Summary: Cluster randomized trials assign clusters of persons to treatments, and all persons in the same cluster receive the same treatment. This paper demonstrates the equivalence of four analysis methods for cluster randomized trials with a pretest and post-test of a quantitative outcome. It also introduces a simple method to calculate the sample size for a cluster randomized trial with baseline measurement.

MULTIVARIATE BEHAVIORAL RESEARCH (2023)

Article Psychology, Clinical

Norms for the Dutch Version of the Young Schema Questionnaire - Adolescent in a Clinical Population

Dyonne M. L. Sijstermans, Marjolijn Span, Gerard J. P. Van Breukelen, Yvonne Bol, Caroline Braet, Jeffrey Roelofs

Summary: The aim of this study was to develop normative data for the five schema domains of the Dutch adolescent version of the Young Schema Questionnaire (YSQ-A). The study also demonstrated the process of obtaining reliable and valid normative data using advanced regression analysis, without the need for subgrouping the norm sample by sex or age but still considering these variables in obtaining norms.

JOURNAL OF PSYCHOPATHOLOGY AND BEHAVIORAL ASSESSMENT (2023)

Article Audiology & Speech-Language Pathology

Pragmatic Uncontrolled Study of Specialized Cognitive Behavioral Therapy for Adults With Chronic Tinnitus

Thomas E. Fuller, Gerard J. P. van Breukelen, Johan W. S. Vlaeyen, Rilana F. F. Cima

Summary: This study examined the effects of stepped-care cognitive behavioral therapy for chronic tinnitus in a regional hospital. The results showed significant improvements in quality of life, tinnitus-related distress, and tinnitus severity among the participants, supporting the implementation of cognitive behavioral therapy in real-world conditions.

EAR AND HEARING (2022)

No Data Available