PLS-SEM-power: A Shiny App and R package for Computing Required Sample Size and Minimum Detectable Effect Size in PLS-SEMs

summary

The gist

PLS-SEM-power: A Shiny App and R package for Computing Required Sample Size and Minimum Detectable Effect Size in PLS-SEMs is introduced to simplify power analysis for Partial Least Squares…

In short

PLS-SEM-power is a Shiny App and R package that simplifies power analysis for Partial Least Squares Structural Equation Modelling (PLS-SEM). It implements Kock and Hadaya's inverse square root method to calculate either the minimum required sample size or the Minimum Detectable Effect Size (MDES). This tool helps researchers plan adequate study designs by determining necessary sample sizes based on desired statistical power and significance levels.

Key concepts

PLS-SEM
Partial Least Squares Structural Equation Modeling is a statistical technique used to analyze complex relationships between latent variables in social and management sciences. It is valued for its flexibility, allowing researchers to manage non-normal data and smaller sample sizes compared to traditional methods.
Inverse Square Root Method
This method, introduced by Kock and Hadaya (2018), determines the minimum required sample size by exploiting the inverse square root relationship between sample size and the standard error of path coefficient estimates. It provides a practical way to find the minimum effect size that can be reliably detected.
Minimum Detectable Effect Size (MDES)
MDES refers to the smallest effect size that a study can reliably detect with a specified level of statistical power. The PLS-SEM-power tool allows researchers to calculate this value given a chosen sample size and significance level, helping them assess the sensitivity of their planned research.

Terminology used across episodes

This episode discusses

The paper

PLS-SEM-power: A Shiny App and R package for Computing Required Sample Size and Minimum Detectable Effect Size in PLS-SEMs · Read on arXiv

Alessandro Ansani, Elena Rinallo

Centre of Excellence in Music, Mind, Body and Brain University of Jyvaskyla · Department of Human Studies Communication Education and Psychology LUMSA University

Despite its evanescent nature, statistical power is crucial for planning Partial Least Squares Structural Equation Modelling (PLS-SEM) studies. This brief paper introduces PLS-SEM-power, a Shiny Application and R package that implements the inverse square root method by Kock and Hadaya (2018) to calculate both the minimum required sample size (a priori analysis) and the Minimum Detectable Effect Size (MDES, sensitivity analysis), given a chosen significance level (alpha level) and desired statistical power (1 - beta), set to 0.80 by default. The software supports both one-tailed and two-tailed calculations, with one-tailed calculations as the default following the formulation of Kock and Hadaya (2018), and provides an intuitive user interface, facilitating reproducible and easily accessible analyses in diverse research contexts.

Transcript

Introduction to the show: ident: Statistics. The day in statistics: methods, theory and the inference under them.

Tom: We're looking at the paper: "PLS-SEM-power: A Shiny App and R package for Computing Required Sample Size and Minimum Detectable Effect Size in PLS-SEMs".

Jane: Here's the gist of the paper. PLS-SEM-power: A Shiny App and R package for Computing Required Sample Size and Minimum Detectable Effect Size in PLS-SEMs is introduced to simplify power analysis for Partial Least Squares Structural Equation Modelling (PLS-SEM) by implementing the inverse square root method from Kock and Hadaya (2018) through a Shiny Application and an R package, which is crucial for planning adequate study designs The interest in Partial Least Squares Structural Equation Models (PLS-SEMs) has been growing exponentially over the last ten years, according to Google Trends and Scopus Several factors explain such growth, including methodological and technological advantages, besides a fundamental philosophical-theoretical difference making the methodology particularly well-suited to a wide range of study situations in social and management science Herman Wold's creation of PLS-SEM in the 1970s provided researchers with an alternative to Covariance-Based Structural Equation Modelling (CB-SEM) Broad acceptance began in the 2000s when accessible software became available, democratizing the technique for researchers without considerable statistical experience PLS-SEMs are valued for their flexibility, managing non-normal data and smaller samples compared to CB-SEM PLS-SEM prioritises prediction over explanation by maximising the variance explained (R2) in dependent constructs Critique of Traditional Sample Size Rules The misconception that sample size computation has no relevance in PLS-SEM stems from the common ‘10-times rule’ (Barclay et al., 1995) which suggests the sample size should be equal to 10 times the number of independent variables in the most complex regression This rule is problematic because it does not account for other critical factors like effect size, significance level (α), and desired statistical power (1−β) The Minimum R-squared method, an alternative approach, assumes researchers have a rough idea of the minimum R2 prior to data collection This method has limitations because it assumes a rough idea of the minimum R2 prior to data collection and predictive endeavor is hardly the main aim in the social sciences The Inverse Square Root Method Kock and Hadaya (2018) introduced the inverse square root method, a versatile tool that determines minimum sample size by exploiting the inverse square root relationship between sample size and standard error of path coefficient estimates The core formula to find the minimum required sample size (N) for a given effect size is N = (pα / pmin)2 Here, pα is a constant depending on the α level, and pmin is the value of the focal path coefficient with the minimum size in the path model that is expected to be statistically significant The rearranged formula expresses pmin as pmin = (pα / √N), which provides a practical approach to determine the minimum effect size that can be reliably detected Implementation and Usage of PLS-SEM-power The PLS-SEM-power Shiny App and R package facilitate these calculations by providing two primary modes of use: A priori (given MDES, compute sample size) and Sensitivity (given sample size, compute MDES) The application supports selections for α level (.01,.05, or.10) and tail selection (one-tailed or two-tailed calculation), with one-tailed calculations selected by default For the R package, functions like pls sem power compute either the minimum sample size needed for a given Minimum Detectable Effect Size (MDES) or the MDES for a given sample size The tool is recommended when the sample size (N) exceeds 10, as smaller samples are better suited for the gamma-exponential method Limitations and Special Considerations The inverse square root method assumes relatively simple model structures and does not account for complexities such as multi-layer constructs or formative indicators Furthermore, the analysis is based on one single path coefficient and not on the size of the most complex regression in the (formative) models It is suggested that researchers should explore a range of MDES when using this method The choice between a one- or two-tailed test should be based on the substantive hypothesis being tested, and PLS-SEM-power provides both options This Shiny App and R package simplify both a priori (sample size estimation given MDES) and sensitivity (MDES estimation given sample size) power analyses for PLS-SEM by interactively implementing the inverse square root method formula from Kock and Hadaya (2018) using one or two-tailed calculations The tool is designed to assist researchers in swiftly evaluating the feasibility of their designs and the detectability of particular effect sizes Thanks to R and Shiny’s open-source nature, this application can be adapted or extended to suit different methodological needs The work was supported by the Research Council of Finland via the Centre of Excellence in Music, Mind, Body and Brain and the FRAMES project

grant numbers 346210 and 368151: (A.

Tom: A.) The authors declared no potential conflicts of interest concerning the research, authorship, and/or publication of this article The PLS-SEM-power Shiny App is freely available at the following link https://aleansani.shinyapps.io/plssem-power/ The R code and a user guide of the plssempower R package are freely available on GitHub at the following link https://github.com/AleAnsani/plssempower The paper was published in 2023 The references section lists numerous foundational works relevant to PLS-SEM and power analysis, including Hair et al.

Jane: (2018) and Kock & Hadaya (2018) The paper is available as arXiv: 2511.14546 Page 7 --- Figure 3.

Tom: Example of a sensitivity power analysis with PLS-SEM-power shows that, with such a sample size, effects greater than 0.30 can be detected with 80% power using the default one-tailed calculation With N = 68 and alpha = 0.05 using a one-tailed test, you can detect effects as small as 0.3 with 80% power The resulting graph indicates that, with such a sample size, effects greater than 0.30 can be detected with 80% power using the default one-tailed calculation Page 9 --- With N = 68 and alpha = 0.05 using a one-tailed test, you can detect effects as small as 0.3 with 80% power The function returns the following message: With N = 68 and alpha = 0.05 using a one-tailed test, you can detect effects as small as 0.3 with 80% power Page 9 --- With N = 68 and alpha =.05 using a one-tailed test, you can detect effects as small as 0.3 with 80% power The function returns the following message: With N = 68 and alpha =.05 using a one-tailed test, you can detect effects as small as 0.3 with 80% power Page 9 --- With N = 68 and alpha =.05 using a one-tailed test, you can detect effects as small as 0.3 with 80% power The function returns the following message: With N = 68 and alpha =.05 using a one-tailed test, you can detect effects as small as 0.

Paper summary: Jane: So, looking back at "PLS-SEM-power: A Shiny App and R package for Computing Required Sample Size and Minimum Detectable Effect Size in PLS-SEMs," what’s the bigger picture here for someone just listening?

Tom: It’s about giving people a practical tool to plan their research designs in PLS-SEM without needing to be a heavy statistical expert just to figure out if they need more data or if their expected effect size is even detectable with their current sample.

Lu: The paper provides an implementation of the inverse square root method from Kock and Hadaya, which is a specific formula used to determine that minimum sample size based on the standard error of path coefficient estimates.

Meng: It makes power analysis for PLS-SEM more accessible by putting it into a Shiny App and an R package, which means you can test out different scenarios quickly in your own environment.

Lalam: The main idea is to simplify both figuring out the required sample size before you start, called a priori analysis, and determining the minimum detectable effect size if you already have a sample size for sensitivity analysis.

Jane: It moves us away from relying solely on rules of thumb, like that ten-times rule mentioned in the paper, toward a calculation that explicitly includes the significance level and power level you choose.

Tom: So, if you're planning a study, this tool helps you check if your sample size is adequate for the effect size you care about under conditions where you want eighty percent power.

Lu: It’s a helpful computational aid for those working in social and management science who are using PLS-SEM and need to justify their sample size choices with more rigorous statistical backing.

Conclusion: Tom: So, we've seen how they built this thing that calculates sample size and effect size for PLS-SEM using this inverse square root method.

Jane: Right, and it turns out these kinds of calculations are super important when you're planning your research design in Partial Least Squares Structural Equation Modelling.

Tom: The title itself is pretty direct: "PLS-SEM-power," and it shows they’ve built both a Shiny App and an R package for this calculation.

Lu: They're taking the old inverse square root method from Kock and Hadaya, which is a way to figure out the minimum sample size based on how much you expect your path coefficients to vary.

Meng: So, it lets you do two main things: first figuring out how big your sample needs to be if you already know what kind of effect size you're looking for.

Jane: And the other thing is sensitivity analysis, where if you have a sample size in mind, it tells you what the smallest effect size your study could possibly detect is.

Tom: It simplifies planning by giving researchers an interactive way to choose things like the significance level and whether they want a one-tailed or two-tailed test.

Lalam: From my side, I see this as a huge win for making these kinds of power analyses accessible to people who aren't deep statistical experts.

Jane: It moves it from being something only PhD students can do on complex software to something that’s much more practical for real researchers in social science and management.

Tom: It addresses the old problem where people used rules like the ten-times rule without actually considering things like the desired power level or effect size.

Lu: The paper's limitation is that this method works best with relatively simple model structures, so if your model has really complex layers or formative indicators, you might need to look at other methods.

Meng: And from a deployment angle, having an R package and a Shiny App means it’s open source, which is good for sharing and adapting it to different research setups.

Jane: It basically gives researchers the power to quickly check the feasibility of their study design before they spend months collecting data that might not be enough.

Tom: So, this tool helps you decide if your planned study is actually going to get the results you need.

More episodes

← Home