metis User Guide
Free, open-source PLS-SEM desktop application for Windows, now available as a public beta with signup-gated access. Built on the seminr R package.
1. What is metis?
metis is a free, open-source desktop application for Partial Least Squares Structural Equation Modelling (PLS-SEM). It is designed to give researchers, academics, and students access to a fully featured PLS-SEM toolkit at no cost — a practical alternative to commercial tools such as SmartPLS for those who cannot afford a subscription.
The application is built on the same statistical engine that powers SmartPLS: the peer-reviewed seminr package for R. The interface is GUI-first — no knowledge of R or programming is required. R runs silently in the background; users interact only with the graphical interface.
2. System Requirements
metis is currently available for Windows only. The following minimum specifications are recommended:
| Requirement | Minimum Specification |
|---|---|
| Operating system | Windows 10 or Windows 11 (64-bit) |
| Processor | 1.6 GHz dual-core (2 GHz+ recommended) |
| RAM | 4 GB (8 GB recommended for large datasets) |
| Disk space | Bundle installer: ~264 MB · Lite installer: ~72 MB |
| Internet | Not required for the current public beta workflow; only needed to submit the signup form or contact support |
| R (Lite only) | R 4.0 or later, already installed system-wide |
3. Choosing Your Installer
metis is distributed as two separate executable files. Both installers provide the full application — the difference is how R is supplied.
metis Bundle
Recommended for most users
- Includes R Portable — no R installation needed
- Fully self-contained, works offline
- seminr pre-installed and configured
- Isolated from any existing R environment
- Ideal for students, first-time users, shared computers
- Larger download due to bundled R runtime
Installation Steps
- Download the bundle installer .exe from the release page.
- Double-click the installer. If Windows Defender SmartScreen shows a warning, click "More info" then "Run anyway" — metis is signed; this warning appears because the certificate is new.
- Follow the on-screen prompts. metis installs to your chosen directory and creates a desktop shortcut.
- On first launch, click Install to extract the R packages. When setup is complete, click Launch to open metis.
metis Lite
For users with R already installed
- Smaller download — faster to install
- Uses your existing system R environment
- seminr installed on first launch automatically
- Ideal for researchers and data analysts already using R
- Does not include R — requires R 4.0+ installed separately
- Download R from cran.r-project.org if needed
Installation Steps
- Ensure R 4.0 or later is installed on your machine. Download from cran.r-project.org if needed.
- Download the lite installer .exe from the release page.
- Double-click the installer and follow the on-screen prompts.
- On first launch, metis detects your system R and prompts you to install the required packages. Click Install and allow the process to complete. When done, click Launch to open metis.
3.3 Comparison at a Glance
| Feature | Bundle (~264 MB) | Lite (~72 MB) |
|---|---|---|
| Includes R runtime | Yes — R Portable | No — uses system R |
| Requires prior R install | No | Yes (R 4.0+) |
| R package setup required | One-time extraction | One-time install |
| Isolated from system R | Yes | No — shared |
| Suitable for non-R users | Yes | No |
| Download size | ~264 MB | ~72 MB |
| Recommended for | Most users | Existing R users |
4. First Launch & Package Setup
Regardless of which installer you used, metis requires a set of R packages to perform statistical calculations. These are installed once on first launch and do not need to be reinstalled unless you reinstall metis or reset your environment.
- seminr — core PLS-SEM engine: estimates models, runs bootstrapping, and computes PLS Predict metrics
- plumber — runs a local REST API on your machine so the metis interface can communicate with R
- semPower — post-hoc power analysis for SEM models
- readxl / jsonlite — data import and result serialisation
No data is sent to any external server during setup or during analysis.
5. Getting Started
5.1 Application Overview
When metis opens, you are taken to the Workspace Home screen. Workspaces are the top-level organising unit in metis — similar to a project folder. Each workspace contains one dataset and one or more models. You can have multiple workspaces for different studies.
5.2 Importing Your Data
- From the Workspace Home, click the "Import File" tab or drag and drop a CSV or Excel (.xlsx) file onto the import area.
- metis will display a delimiter and encoding configuration screen. Confirm your settings (comma-separated is the default for CSV files).
- You will then see a descriptive statistics table showing means, standard deviations, skewness, kurtosis, and scale information for all variables. Review this before proceeding.
- Click "Continue to Model Builder" to proceed to the canvas.
5.3 Building Your Model
- The Model Canvas is the central workspace. Use the toolbar to add Latent Variables (constructs) and draw paths between them.
- Drag variables from the left-hand panel onto constructs to assign indicators.
- Set each construct as Reflective or Formative using the Properties panel on the right.
- Path arrows represent structural relationships. Draw them from predictor to outcome constructs.
- Your model is saved automatically every 5 minutes. You can also save manually using the Save button.
5.4 Running the Analysis
- Click the "Calculate" button in the toolbar to run the PLS-SEM algorithm.
- A settings dialogue will appear. Configure the algorithm (Standard PLS or Consistent PLS), weighting mode, and stopping criterion.
- Once confirmed, metis runs the estimation in the background using the seminr R engine. Results appear in the Results View automatically.
5.5 Interpreting Results
The Results View displays a full SmartPLS-style results tree on the left and the path diagram on the right. The following result sections are available:
| Section | Contents |
|---|---|
| Final Results | Total indirect effects, specific indirect effects, total effects, outer loadings, outer weights, latent variables, residuals |
| Quality Criteria | R², f², construct reliability and validity, discriminant validity (HTMT & Fornell-Larcker), collinearity (VIF), model fit, model selection criteria (AIC/BIC) |
| Algorithm | Settings used, stop criterion changes, post-hoc power analysis (via semPower) |
| Model and Data | Inner model, outer model, indicator data (original / standardised / correlations), measurement model |
| Bootstrap Results | Path coefficients with confidence intervals, t-values, and p-values |
| PLS Predict | Out-of-sample Q² predict, RMSE comparisons (PLS-SEM vs. LM benchmark) |
5.6 Bootstrap Analysis
Bootstrapping is the recommended method for testing the significance of path coefficients in PLS-SEM.
- From the Model Canvas, select Analysis › Bootstrap, or click the Bootstrap button in the toolbar.
- Set the number of subsamples (default 5,000), confidence interval type, and confidence level.
- Click Run. A progress indicator will appear; bootstrapping cannot be interrupted once started.
- Bootstrap results are saved as a separate entry in your workspace and can be reopened at any time without re-running.
5.7 PLS Predict
PLS Predict evaluates the out-of-sample predictive power of your model using k-fold cross-validation. Results include Q² predict values and RMSE metrics compared to a linear model (LM) benchmark, following the SmartPLS 4 reporting convention.
- Select Analysis › PLS Predict from the menu bar, or click the PLS Predict button.
- Configure the number of folds and repetitions in the settings dialogue.
- Results show Q²predict, PLS-SEM RMSE, and LM RMSE for each endogenous indicator.
5.8 Exporting Results
Generates a shareable single-file HTML report with a fixed sidebar for navigation and scrollspy highlighting. Opens automatically in your default browser.
Copies the equivalent seminr R code to your clipboard so you can reproduce the analysis in a standalone R session.
6. Data Privacy & Security
metis is designed with privacy as a foundational principle, not an afterthought. This is particularly important for researchers working with human subjects data.
7. Known Limitations — v0.1.7
- Windows only. Linux and macOS support is planned for Phase 2, subject to sufficient user interest.
- Public beta scope. AI-assisted reporting is not included in the current public beta release.
- Stop criterion history. Detailed stop-criterion change history across PLS algorithm iterations is not available as a direct output from the seminr backend. A metadata summary is provided instead.
- Large datasets. Very large datasets (over 10,000 rows) may cause slower bootstrap calculation times depending on hardware.
- Execution logs. Execution logs are not currently exportable as a standalone file.
Licence & Support
metis is released under the GNU General Public Licence v3 (GPL-3). You are free to use, study, modify, and distribute the application in accordance with the terms of that licence.
For questions, bug reports, or feature requests, please open an issue on the metis GitHub repository. Community contributions are welcome.
Kwame Nkrumah University of Science and Technology (KNUST), Ghana.
Department of Educational Innovation in Science and Technology.
View all team members
Supervised by: Prof Harry Barton Essel and Dr Akosua Techie-Menson.