GLMM · Mixed Logistic Regression · Random Effects
Mixed Logistic Regression Calculator — Free GLMM Tool
A free online generalized linear mixed model (GLMM) calculator for binary outcomes with clustered or repeated-measures data. Compute fixed effects, random-intercept variance, odds ratios, ICC, AIC/BIC, and APA-format results — instantly.
Logistic Curve
Random Intercepts
Odds Ratio Forest
ICC · Variance
📊 Enter Your Binary Outcome Data
Enter the binary outcome (0/1) for each cluster (e.g., school, hospital, patient ID). Each cluster is one row of comma-separated values. Random intercepts will be estimated for each cluster.
Dataset auto-loads on selection.
Leave empty to fit an intercept-only random model.
Headers auto-detected. Each column = one cluster. Values must be 0 or 1 for binary outcome.
Use the Paste / Type tab above — each cluster has its own editable name and binary 0/1 series. Add or remove clusters with the buttons.
Laplace approximation, 60 default.
🔍 Detailed Interpretation of Results & How to Write Your Results in Research
Subsection 1 — Detailed Interpretation Results
A plain-language, paragraph-by-paragraph interpretation of every part of the mixed logistic regression output, written for non-statisticians and updated dynamically with the current numbers.
Run the analysis above to see a 6-paragraph guided interpretation of the fixed effects, odds ratio, random-intercept variance, ICC, statistical and practical significance, and study limitations.
Subsection 2 — How to Write Your Results in Research (5 Examples)
Five ready-to-use templates, each auto-filled with your current statistics and one-click copyable. Use the style that matches your audience.
📝 Conclusion — What This Mixed Logistic Regression Tells You
Run the analysis above to generate a fully written conclusion section that you can copy directly into a thesis, dissertation chapter, or journal manuscript.
🎯 When to Use Mixed Logistic Regression
This free mixed logistic regression calculator is designed for binary outcomes (0/1) collected from clustered or repeated-measures data. Use it whenever observations within a group cannot be assumed independent — a common feature of educational, medical, ecological, and organisational research where the p-value from a standard logistic regression would be misleadingly small.
Use this tool when…
- Your outcome is binary (success/failure, yes/no, detected/not detected, alive/dead, recovered/not recovered).
- Observations are nested in groups (students in schools, patients in hospitals, animals in reserves, customers in stores).
- You measured the same subject more than once (repeated measures, longitudinal binary outcomes).
- You expect an intraclass correlation (ICC) above 0.05, indicating meaningful between-cluster variability.
- You need population-average odds ratios with valid standard errors that account for clustering.
- You want to quantify how much of the variance in the outcome lies between clusters versus within.
- You want to report results in APA 7th, thesis, structured-abstract, plain-language, or pre-registration format.
Concrete worked examples
- Education: Did a new teaching method increase the probability of passing a national exam, after accounting for clustering of students within schools?
- Healthcare: Does a 30-day hospital readmission depend on a treatment, with patients clustered within hospitals?
- Wildlife ecology: Does habitat type predict camera-trap detection of a focal species, with cameras clustered within reserves?
- Marketing: Does a loyalty programme raise the probability of a repeat purchase, with customers clustered within stores?
Decision tree
Is the outcome binary (0/1)?
├── No → Use Linear Mixed Model (continuous), Mixed Poisson (counts), or Mixed Cox (survival)
└── Yes
└── Are observations clustered or repeated?
├── No → Use ordinary Logistic Regression (GLM)
└── Yes → Use MIXED LOGISTIC REGRESSION (GLMM) ← this tool
📐 Technical Notes — Formulas & Estimation
Model
The two-level random-intercept binary GLMM is:
logit(Pᵢⱼ) = log( Pᵢⱼ / (1 − Pᵢⱼ) ) = β₀ + β₁ Xⱼ + uⱼ
uⱼ ~ N(0, σ²ᵤ)
Yᵢⱼ ~ Bernoulli(Pᵢⱼ)
Where: Yᵢⱼ = the binary outcome of observation i in cluster j; Pᵢⱼ = probability that Yᵢⱼ = 1; β₀ = overall log-odds intercept; β₁ = fixed slope of the cluster-level predictor X; uⱼ = cluster-specific random intercept; σ²ᵤ = between-cluster variance on the log-odds scale.
Odds Ratio
OR = exp(β) 95% CI = exp(β ± z_{α/2} · SE(β))
Intraclass Correlation Coefficient (latent-variable formulation)
ICC = σ²ᵤ / (σ²ᵤ + π² / 3) where π² / 3 ≈ 3.2899
Estimation
This calculator uses an iteratively-reweighted least-squares Laplace approximation for the random intercepts and a Newton–Raphson update for fixed effects, with REML-style variance shrinkage. For very large datasets or random slopes, software such as R lme4::glmer() with adaptive Gauss–Hermite quadrature is recommended.
Model Fit Indices
Deviance = −2 · log L
AIC = Deviance + 2 · k
BIC = Deviance + k · log(N)
McFadden R² = 1 − (log L_full / log L_null)
📚 How to Use This Mixed Logistic Regression Calculator — Step-by-Step
- Enter Your Data: Use the Paste / Type tab. Each cluster (school, hospital, patient) gets its own row with a binary 0/1 series. The placeholder shows the format
52, 48, 55, 61, 47, …— values can be raw counts that the tool dichotomises by median, or already-binary 0/1 entries. The cluster name is editable; rename it to match your research. - Choose a Sample Dataset: Five built-in datasets (hospitals, schools, patients, stores, wildlife reserves) load instantly. Use them to learn the workflow before entering your own.
- Configure Test Settings: Pick alpha (0.01, 0.05, or 0.10). Choose Random intercept only for an unconditional model or Random intercept + predictor if you have a cluster-level X. Iterations (default 60) usually do not need changing.
- Run the Analysis: Click the green 🚀 Run Mixed Logistic Regression button. Results appear in under a second.
- Read the Summary Cards: Four colour-coded cards show the overall log-odds intercept, slope, odds ratio, and ICC. Green = significant at α; amber = borderline; red = non-significant.
- Read the Full Results Tables: The Fixed Effects table reports B, SE, OR, 95% CI, z, and p for every term. The Random Effects table reports σ²ᵤ and the ICC.
- Examine Both Visualizations: Chart 1 shows cluster-level proportions vs the predictor with a fitted random-intercepts line per cluster. Chart 2 shows the population-averaged predicted probability curve plus an ICC variance partition donut.
- Check Assumptions: Independence, binomial response, normality of random effects, linearity of the logit, and adequate cluster count are flagged automatically with green/amber/red badges in the interpretation section.
- Read the Interpretation: The 6-paragraph interpretation translates B, OR, σ²ᵤ, and ICC into plain English you can paste straight into a discussion section.
- Export Your Results: Use Download Doc for a plain-text .txt report, or Download PDF for an A4 publication-ready PDF via your browser's print dialog.
❓ Frequently Asked Questions
Q1. What is mixed logistic regression and when should I use it?
Mixed logistic regression — a binary generalized linear mixed model (GLMM) — extends standard logistic regression by adding random effects that account for clustering. Use it whenever a binary outcome (0/1) comes from observations grouped within higher-level units such as patients within hospitals, students within schools, or repeated measures within the same person.
Q2. What is the difference between logistic regression and mixed logistic regression?
Standard logistic regression assumes independence of all observations. Mixed logistic regression relaxes this by introducing a random intercept (and optionally random slopes) for each cluster, which models the within-cluster correlation and produces valid standard errors. Ignoring clustering in standard logistic regression typically inflates Type I error rates substantially.
Q3. How do I interpret the odds ratio in a mixed logistic regression?
An odds ratio (OR) of 2.0 means that, holding the cluster random effect constant, the odds of the outcome (Y = 1) are twice as high for each one-unit increase in the predictor. OR > 1 indicates a positive effect, OR < 1 a protective effect, and OR ≈ 1 no effect. Always report the 95% confidence interval alongside the point estimate, since wide intervals signal weak evidence even when the OR appears large.
Q4. What does ICC mean in a binary GLMM?
The intraclass correlation coefficient measures the share of total variance attributable to between-cluster differences on the latent log-odds scale. It is computed as σ²ᵤ / (σ²ᵤ + π²/3) ≈ σ²ᵤ / (σ²ᵤ + 3.29). An ICC above 0.05 typically justifies a mixed model; an ICC above 0.20 indicates strong clustering that would severely bias a standard logistic regression.
Q5. What are random intercepts versus random slopes?
A random intercept allows each cluster to have its own baseline log-odds. A random slope additionally lets the effect of a predictor vary across clusters. Most applied analyses begin with random intercepts only; add random slopes only when a likelihood-ratio test or theoretical argument supports cluster-specific effects, since random slopes triple the number of variance parameters.
Q6. How large a sample do I need for mixed logistic regression to be reliable?
A common rule of thumb is at least 30 clusters with at least 5 observations per cluster, plus at least 10 events (Y = 1) per fixed-effect predictor. Below 30 clusters, random-effect variance estimates become unstable; below 10 events per predictor, the fixed effects become unreliable. For small studies, consider penalised likelihood (e.g., logistf) or a Bayesian GLMM with weakly informative priors.
Q7. What does a non-significant random-effect variance mean?
If σ²ᵤ is near zero and its likelihood-ratio test is non-significant, observations within a cluster behave like independent observations and a standard logistic regression may be sufficient. However, do not drop the random effect on the basis of a single test alone — keeping it protects against under-estimating standard errors and is the more conservative inferential choice.
Q8. How do I report mixed logistic regression results in APA 7th edition format?
Report the fixed-effect estimate (B), standard error, odds ratio with 95% confidence interval, z-value, and p-value, plus the random-effect variance σ²ᵤ and the ICC. Example: "B = 0.84, SE = 0.21, OR = 2.32, 95% CI [1.54, 3.49], z = 4.00, p < .001; σ²ᵤ = 0.42, ICC = 0.11." See the five reporting templates in the section above for full sentence-level wording.
Q9. Does mixed logistic regression assume normality of the outcome?
No. The outcome is binary and follows a Bernoulli distribution. The normality assumption applies to the random effects (cluster intercepts) on the log-odds scale, not to the response itself. Independence within clusters is replaced by conditional independence given the random effect.
Q10. Can I use this calculator for my published research or thesis?
Yes for exploratory analysis, teaching, coursework, and methodological learning. For peer-reviewed publication, verify the results with R
lme4::glmer(), SAS PROC GLIMMIX, or Stata melogit. Cite the tool as: Stats Unlock. (2025). Mixed logistic regression calculator. https://statsunlock.com.📚 References
The following references underpin the design of this mixed logistic regression calculator and inform its formulas, p-value computation, and effect-size reporting. They cover both the theory of generalized linear mixed models and best practice for multilevel logistic regression in applied research.
- Agresti, A. (2015). Foundations of linear and generalized linear models. Wiley. View book
- Bates, D., Mächler, M., Bolker, B., & Walker, S. (2015). Fitting linear mixed-effects models using lme4. Journal of Statistical Software, 67(1), 1–48. https://doi.org/10.18637/jss.v067.i01
- Bolker, B. M., Brooks, M. E., Clark, C. J., Geange, S. W., Poulsen, J. R., Stevens, M. H. H., & White, J. S. S. (2009). Generalized linear mixed models: A practical guide for ecology and evolution. Trends in Ecology & Evolution, 24(3), 127–135. https://doi.org/10.1016/j.tree.2008.10.008
- Breslow, N. E., & Clayton, D. G. (1993). Approximate inference in generalized linear mixed models. Journal of the American Statistical Association, 88(421), 9–25. https://doi.org/10.1080/01621459.1993.10594284
- Cohen, J. (1988). Statistical power analysis for the behavioral sciences (2nd ed.). Lawrence Erlbaum. https://doi.org/10.4324/9780203771587
- Goldstein, H. (2011). Multilevel statistical models (4th ed.). Wiley. https://doi.org/10.1002/9780470973394
- Hedeker, D. (2003). A mixed-effects multinomial logistic regression model. Statistics in Medicine, 22(9), 1433–1446. https://doi.org/10.1002/sim.1522
- Hox, J. J., Moerbeek, M., & Van de Schoot, R. (2017). Multilevel analysis: Techniques and applications (3rd ed.). Routledge. https://doi.org/10.4324/9781315650982
- McCulloch, C. E., Searle, S. R., & Neuhaus, J. M. (2008). Generalized, linear, and mixed models (2nd ed.). Wiley. View book
- Nakagawa, S., & Schielzeth, H. (2013). A general and simple method for obtaining R² from generalized linear mixed-effects models. Methods in Ecology and Evolution, 4(2), 133–142. https://doi.org/10.1111/j.2041-210x.2012.00261.x
- Raudenbush, S. W., & Bryk, A. S. (2002). Hierarchical linear models: Applications and data analysis methods (2nd ed.). Sage. View book
- Snijders, T. A. B., & Bosker, R. J. (2012). Multilevel analysis: An introduction to basic and advanced multilevel modeling (2nd ed.). Sage. View book
- Stroup, W. W. (2013). Generalized linear mixed models: Modern concepts, methods and applications. CRC Press. https://doi.org/10.1201/b13151
- Wickham, H. (2016). ggplot2: Elegant graphics for data analysis. Springer-Verlag. https://doi.org/10.1007/978-3-319-24277-4
- Zuur, A. F., Ieno, E. N., Walker, N. J., Saveliev, A. A., & Smith, G. M. (2009). Mixed effects models and extensions in ecology with R. Springer. https://doi.org/10.1007/978-0-387-87458-6
