sample size planning
Recently Published Documents


TOTAL DOCUMENTS

69
(FIVE YEARS 21)

H-INDEX

12
(FIVE YEARS 1)

2021 ◽  
pp. 096228022110510
Author(s):  
Stefan Wellek

More often than not, clinical trials and even nonclinical medical experiments have to be run with observational units sampled from populations to be assumed heterogeneous with respect to covariates associated with the outcome. Relevant covariates which are known prior to randomization are usually categorical in type, and the corresponding subpopulations are called strata. In contrast to randomization which in most cases is performed in a way ensuring approximately constant sample size ratios across the strata, sample size planning is rarely done taking stratification into account. This holds true although the statistical literature provides a reasonably rich repertoire of testing procedures for stratified comparisons between two treatments in a parallel group design. For all of them, at least approximate methods of power calculation are available from which algorithms or even closed-form formulae for required sample sizes can be derived. The objective of this tutorial is to give a systematic review of the most frequently applicable of these methods and to compare them in terms of their efficiency under standard settings. Based on the results, recommendations for the sample size planning of stratified two-arm trials are given.


2021 ◽  
Author(s):  
Samuel Donnelly ◽  
Terrence D. Jorgensen ◽  
Cort Rudolph

Conceptual and statistical models that include conditional indirect effects (i.e., so-called “moderated mediation” models) are increasingly popular in the behavioral sciences. Although there is ample guidance in the literature for how to specify and test such models, there is scant advice regarding how to best design studies for such purposes, and this especially includes techniques for sample size planning (i.e., “power analysis”). In this paper, we discuss challenges in sample size planning for moderated mediation models and offer a tutorial for conducting Monte Carlo simulations in the specific case where one has categorical exogenous variables. Such a scenario is commonly faced when one is considering testing conditional indirect effects in experimental research, wherein the (assumed) predictor and moderator variables are manipulated factors and the (assumed) mediator and outcome variables are observed/measured variables. To support this effort, we offer example data and reproducible R code that constitutes a “toolkit” to aid researchers in the design of research to test moderated mediation models.


Author(s):  
Joseph P. Vitta ◽  
Christopher Nicklin ◽  
Stuart McLean

Abstract In this focused methodological synthesis, the sample construction procedures of 110 second language (L2) instructed vocabulary interventions were assessed in relation to effect size–driven sample-size planning, randomization, and multisite usage. These three areas were investigated because inferential testing makes better generalizations when researchers consider them during the sample construction process. Only nine reports used effect sizes to plan or justify sample sizes in any fashion, with only one engaging in an a priori power procedure referencing vocabulary-centric effect sizes from previous research. Randomized assignment was observed in 56% of the reports while no report involved randomized sampling. Approximately 15% of the samples observed were constructed from multiple sites and none of these empirically investigated the effect of site clustering. Leveraging the synthesized findings, we conclude by offering suggestions for future L2 instructed vocabulary researchers to consider a priori effect size–driven sample planning processes, randomization, and multisite usage when constructing samples.


2021 ◽  
Author(s):  
Björn Jörges

Sample size planning is not straight-forward for the complex designs that are usually employed in psychophysical (two-alternative forced-choice) experiments: characteristics such as binary response variables and nested data structures where responses may be correlated differently within participants and experimental sessions than across participants and experimental sessions make it harder to estimate the necessary number of participants and trials with traditional means. In this practical R-based guide, we first show in detail how we can simulate verisimilar psychophysical data. We then use these simulations to compare two different methods by which two-alternative forced-choice data can be analyzed: (1) the “two-step” approach, where first psychometric functions are fitted and then statistical tests are performed over the parameters of these fitted psychometric functions; (2) an approach based on Generalized Linear Mixed Modeling (GLMM) that does not require the intermediary step of fitting psychometric functions. We argue that the GLMM approach enhances statistical validity and show that it can increase statistical power. Finally, we provide a sample implementation of a simulation-based power analysis that can be used as-is for many simple designs, but is also easily adaptable for more complex designs. Overall, we show that a GLMM-based approach can be beneficial for data analysis and sample size planning for typical (two-alternative forced-choice) psychophysical designs.


Psychometrika ◽  
2021 ◽  
Author(s):  
Gwowen Shieh

A Correction to this paper has been published: https://doi.org/10.1007/s11336-019-09692-3


2021 ◽  
pp. 1-7
Author(s):  
Raphael Schuster ◽  
Tim Kaiser ◽  
Yannik Terhorst ◽  
Eva Maria Messner ◽  
Lucia-Maria Strohmeier ◽  
...  

Abstract Background Sample size planning (SSP) is vital for efficient studies that yield reliable outcomes. Hence, guidelines, emphasize the importance of SSP. The present study investigates the practice of SSP in current trials for depression. Methods Seventy-eight randomized controlled trials published between 2013 and 2017 were examined. Impact of study design (e.g. number of randomized conditions) and study context (e.g. funding) on sample size was analyzed using multiple regression. Results Overall, sample size during pre-registration, during SSP, and in published articles was highly correlated (r's ≥ 0.887). Simultaneously, only 7–18% of explained variance related to study design (p = 0.055–0.155). This proportion increased to 30–42% by adding study context (p = 0.002–0.005). The median sample size was N = 106, with higher numbers for internet interventions (N = 181; p = 0.021) compared to face-to-face therapy. In total, 59% of studies included SSP, with 28% providing basic determinants and 8–10% providing information for comprehensible SSP. Expected effect sizes exhibited a sharp peak at d = 0.5. Depending on the definition, 10.2–20.4% implemented intense assessment to improve statistical power. Conclusions Findings suggest that investigators achieve their determined sample size and pre-registration rates are increasing. During study planning, however, study context appears more important than study design. Study context, therefore, needs to be emphasized in the present discussion, as it can help understand the relatively stable trial numbers of the past decades. Acknowledging this situation, indications exist that digital psychiatry (e.g. Internet interventions or intense assessment) can help to mitigate the challenge of underpowered studies. The article includes a short guide for efficient study planning.


2020 ◽  
Author(s):  
Julian Schuessler ◽  
Markus Freitag

Conjoint experiments aiming to estimate average marginal component effects and related quantities have become a standard tool for social scientists. However, existing solutions for power analyses to find appropriate sample sizes for such studies have various shortcomings and accordingly, explicit sample size planning is rare. Based on recent advances in statistical inference for factorial experiments, we derive simple yet generally applicable formulae to calculate power and minimum required sample sizes for testing average marginal component effects (AMCEs), conditional AMCEs, as well as interaction effects in forced-choice conjoint experiments. The only input needed are expected effect sizes. Our approach only assumes random sampling of individuals or randomization of profiles and avoids any parametric assumption. Furthermore, we show that clustering standard errors on individuals is not necessary and does not affect power. Our results caution against designing conjoint experiments with small sample sizes, especially for detecting heterogeneity and interactions. We provide an R package that implements our approach.


Sign in / Sign up

Export Citation Format

Share Document