Commercial SEM packages are expensive, and syntax-based tools have a steep learning curve. If you just need to test a research model for a thesis or paper, you can now do the whole analysis in your browser for free with CB-SEM Studio: draw the model with your mouse, click Run, and get fit indices, reliability and validity tables, path coefficients, mediation and moderation results, ready to paste into your report.
This tutorial walks through a complete analysis using the built-in sample study, so you can follow along without your own data.
In this guide
- What you need
- Step 1: Load your data
- Step 2: Build the measurement model
- Step 3: Add the structural paths
- Step 4: Add moderation (optional)
- Step 5: Run the analysis
- Step 6: Check model fit
- Step 7: Evaluate the measurement model
- Step 8: Interpret the structural model
- Step 9: Improve fit responsibly with modification indices
- Step 10: Export and cite
- CB-SEM Studio vs commercial SEM software
- Frequently asked questions
What you need
- Item-level survey data: one row per respondent and one column per questionnaire item (for example PE1, PE2, PE3…). CSV, TSV and Excel files all work.
- A research model: which items measure which construct, and which constructs you expect to affect which.
- A modern browser. Your data are processed on your own device and never uploaded.
If your data need recoding first (reverse-scored items, text labels such as "Agree" → 4, missing value codes), clean them with the free data cleaning and coding tool and send them straight to the studio.
Step 1: Load your data
Open CB-SEM Studio and drop your file onto the Data tab, or click Sample study to load a 300-respondent technology-acceptance survey. Choose how to handle missing values: listwise deletion or item-mean replacement. Check the variable list for impossible values or wrongly coded items before moving on.
Step 2: Build the measurement model
On the Model tab:
- Click + New construct, name it (for example "Performance expectancy"), and add its items from the list.
- Repeat for every construct.
- Drag constructs around the canvas to lay out your model, and use the item side dropdown on each construct card to place its items left, right, above or below. Click Tidy layout for an automatic arrangement.
- Add or remove items on each construct card at any time.
With no arrows between constructs, the model is a confirmatory factor analysis (CFA): every construct correlates freely with every other. Running it first is good practice, because it lets you check the measurement model before testing hypotheses. Our CFA tutorial explains each output.
Step 3: Add the structural paths
Pick a From and a To construct in the Paths box and click Add, or switch to the Draw path tool and drag an arrow from one construct to another. The Covariance tool adds correlations, for example between two exogenous constructs or two error terms. Erase removes anything you click.
Mediation needs no extra setup. When you draw X → M → Y, the studio finds every indirect effect automatically and estimates it with Monte Carlo confidence intervals.
Step 4: Add moderation (optional)
In the Moderation box, choose the predictor, the moderator and the outcome. The studio builds the latent interaction for you (with product indicators, double-mean-centered), adds it to the diagram as a dashed "X × M" construct, and later reports simple slopes at ±1 SD of the moderator, with a slope plot. If a moderated path is also part of an indirect effect, you get the index of moderated mediation too.
Step 5: Run the analysis
On the Run tab, keep the defaults (maximum likelihood, first loading fixed to 1, α = .05) and click Run analysis. Estimation takes a few seconds.
Step 6: Check model fit
The Model fit tab compares the fit indices with common cut-offs (Hu & Bentler, 1999):
| Index | Good fit | Acceptable |
|---|---|---|
| CFI / TLI | ≥ .95 | ≥ .90 |
| RMSEA | ≤ .06 | ≤ .08 |
| SRMR | ≤ .08 | ≤ .10 |
| χ²/df | ≤ 3 | ≤ 5 |
The chi-square test is almost always significant with samples of a few hundred, so judge fit mainly on CFI, TLI, RMSEA and SRMR. See model fit indices explained for details.
Step 7: Evaluate the measurement model
The Measurement model tab gives:
- standardized loadings (ideally ≥ .70);
- Cronbach's alpha, composite reliability (≥ .70) and AVE (≥ .50);
- discriminant validity with HTMT (< .85) and the Fornell-Larcker table.
If something fails, see our guides to Cronbach's alpha and discriminant validity.
Step 8: Interpret the structural model
The Structural model tab shows unstandardized and standardized path coefficients, standard errors, z-values, p-values and the R² of every endogenous construct. Write each hypothesis as supported or not supported based on the sign and significance of its path.
The Mediation tab lists every indirect effect, with its Monte Carlo CI and total effects. The Moderation tab gives the interaction effect, simple slopes and plots.
Step 9: Improve fit responsibly with modification indices
If fit is poor, the Modification indices tab lists the parameters that would most reduce χ², with the expected parameter change. Each one has an Add & rerun button. Only add parameters that make theoretical sense, such as correlated errors between two items with near-identical wording. Report every change, because chasing fit with modification indices capitalizes on chance.
Step 10: Export and cite
The Export tab saves your results as an Excel workbook, a Word document, HTML or LaTeX tables, the diagram as SVG or a 300-dpi PNG, and the model as lavaan syntax, so you can reproduce it in R if a reviewer asks. It also writes a methods paragraph you can adapt. Save the whole analysis as a project file to reopen later.
If CB-SEM Studio helped with your thesis or paper, please cite it. The citation is on the Export tab:
Xavier, B. (2026). CB-SEM Studio: Free covariance-based structural equation modeling software [Web application]. FreeStatisticalTools.com. https://freestatisticaltools.com/cb-sem/
CB-SEM Studio vs commercial SEM software
| CB-SEM Studio | Typical commercial package | |
|---|---|---|
| Price | Free | Paid licence |
| Installation | None (runs in browser) | Desktop install, often Windows-only |
| Model building | Drag and drop, dropdowns | Drawing or syntax |
| Estimation | Maximum likelihood | ML and others |
| Reliability and validity tables | Built in (alpha, CR, AVE, HTMT, Fornell-Larcker) | Often calculated by hand or with plug-ins |
| Latent moderation | Built in, with simple slopes | Varies |
| Data privacy | Data stay on your device | Local |
For very advanced needs (multi-group invariance testing, categorical indicators with WLSMV, multilevel SEM), a dedicated package such as lavaan or Mplus is still the right tool. You can export your model as lavaan syntax to get started.
If your goal is prediction or your model has formative constructs, consider PLS-SEM instead. Our CB-SEM vs PLS-SEM comparison explains the difference, and PLS-SEM Studio is free too.
Frequently asked questions
Can I run SEM without AMOS or R?
Yes. CB-SEM Studio runs covariance-based SEM and CFA by maximum likelihood in your web browser, with a drag-and-drop model builder. No installation or account is needed.
Is my data uploaded to a server?
No. The analysis runs entirely in your browser, so your data never leave your device.
Does it report the fit indices reviewers ask for?
Yes: chi-square, df, χ²/df, CFI, TLI, RMSEA with its 90% confidence interval, SRMR, and more, together with reliability and validity tables.
Can it test mediation and moderation?
Yes. Indirect effects are found automatically from your paths and tested with Monte Carlo confidence intervals. Latent moderation is added from a dropdown and comes with simple slopes and the index of moderated mediation.
How should I cite CB-SEM Studio?
Use: Xavier, B. (2026). CB-SEM Studio: Free covariance-based structural equation modeling software [Web application]. FreeStatisticalTools.com. https://freestatisticaltools.com/cb-sem/