Overview
FlexibleGLMM is an open-source R Shiny application designed to provide an interactive graphical user interface (GUI) for performing Generalized Linear Mixed Model (GLMM) analyses without requiring extensive programming knowledge. The application integrates multiple established R packages, including afex, lme4, nlme, emmeans, and a broad ecosystem of diagnostic and reporting packages, into a unified workflow covering:
- Data preprocessing and quality control
- Variable selection and standardization
- Missing data handling
- Outlier detection and removal
- Distribution fitting for selecting appropriate model families
- Generalized linear mixed model fitting
- SAS PROC MIXED/PROC GLIMMIX-style Type III hypothesis testing
- Estimated marginal means (least-square means) and post-hoc comparisons
- Model diagnostics and performance assessment
- Publication-ready tables and visualizations
- Automated reproducible reports
The application is particularly useful for researchers in biomedical, psychological, clinical, and social sciences who routinely analyze repeated-measures, longitudinal, and hierarchical datasets.
Motivation
Mixed-effects modeling has become the standard approach for analyzing repeated measurements and clustered data. However, performing a complete GLMM analysis often requires combining several R packages and writing complex statistical code. Many researchers continue to to rely on different and at times multiple statistical tools. Translating analyses between them and R often requires substantial programming effort and familiarity with differences in covariance structures, denominator degrees-of-freedom approximations, hypothesis testing procedures, and least-squares means estimation. FlexibleGLMM was developed to address these challenges by bringing together the flexibility of the R statistical ecosystem with a user-friendly point-and-click interface that allows users to conduct sophisticated mixed-model analyses and obtain results comparable to traditional SAS/SPSS workflows.
Web Application
FlexibleGLMM can be accessed directly without installation through the Shiny web application:
Live Application:
https://puneet-talwar.shinyapps.io/FlexibleGLMM/
Installation:
- Install devtools if not already installed
install.packages("devtools")
library(devtools)
- Install FlexibleGLMM package
devtools::install_github("puneettalwar/FlexibleGLMM")
Launch the App
Either with
library(FlexibleGLMM)
run_app()
or
FlexibleGLMM::run_app()
Source Code
The complete source code is available on GitHub:
Repository:
https://github.com/puneettalwar/FlexibleGLMM
Users are welcome to report bugs, suggest new features, and contribute to future development through GitHub Issues and Pull Requests.
Application Workflow
FlexibleGLMM follows a sequential workflow from raw data preprocessing to final reporting.
Data Import
↓
Data Cleaning & Preprocessing
↓
Distribution Assessment
↓
Model Specification
↓
GLMM Model Fitting
↓
ANOVA & Type III Tests
↓
Estimated Marginal Means
↓
Model Diagnostics
↓
Visualization & Reporting
Main Interface
The application consists of a left sidebar for configuring the analysis and a main panel containing results tabs.
Sidebar Features
1. Data Upload and Variable Selection
- Import datasets in .csv and .xlsx formats.
- Automatically detect variables.
- Select variables to include or exclude from analysis.
- Define variable types (continuous or categorical).
2. Missing Value Treatment
The application provides options for handling incomplete observations:
- Detect missing entries.
- Remove rows containing missing values.
- Export the cleaned dataset for external use.
3. Outlier Detection and Removal
Multiple outlier detection approaches are available:
Z-score Method:
- Identify observations exceeding a user-defined standard deviation cutoff.
Cook’s Distance:
- Detect influential observations using the common threshold: [4/n]
Mahalanobis Distance:
- Detects multivariate outliers based on a user-selected chi-square quantile threshold.
Lookout:
- Lookout provides an alternative approach for detecting unusual observations using leave-one-out kernel density estimates and extreme value theory.
DHARMa:
- Detects outliers using simulated model residuals; therefore, a fitted model is required before diagnostic plots can be generated.
Detected observations can be removed and the cleaned dataset downloaded.
4. Numerical Data Standardization
Continuous variables can be transformed using:
- Centering and scaling (standardization)
- Centering only
- Scaling only
- No transformation The processed data can be viewed directly within the application (in the Data tab).
5. Distribution Fitting
For continuous dependent variables, FlexibleGLMM can compare candidate probability distributions using goodness-of-fit statistics. This helps users select an appropriate distribution prior to model fitting.
6. Interaction Specification
Users can automatically specify:
- No interaction
- Two-way interactions
- Three-way interactions for investigating complex relationships among predictors.
7. GLMM Family and Link Functions
The application supports selection of:
- Response distribution family (gaussian, gamma, beta, binomial and poisson)
- Appropriate link functions (default, identity, log, logit, probit, cloglog, sqrt, inverse) allowing analysis of both Gaussian and non-Gaussian outcomes.
8. Modeling Engines
FlexibleGLMM integrates multiple mixed-model frameworks. afex::mixed
Provides:
- Type III ANOVA
- SAS-like hypothesis testing
- Support for complex factorial designs
- Appropriate denominator degree-of-freedom approximations
lme4
Provides:
- Linear mixed models
- Generalized linear mixed models
- Flexible random-effects specification
nlme
Provides:
- Linear mixed models with explicit residual correlation structures
- Repeated-measures covariance modeling
9. Correlation Structures
For nlme models, users can specify correlation structures commonly required in longitudinal studies, including:
- Compound symmetry
- Autoregressive AR(1)
- Other supported correlation structures Additional fields allow users to define:
- Subject/grouping variables
- Time variables for autoregressive models
10. Random Effects Specification
FlexibleGLMM supports both lme4/afex and nlme random-effect syntax. Users can specify:
- Random intercepts
- Random slopes
- Nested random effects
- Crossed random effects
- Correlated and uncorrelated random structures The application includes a reference table illustrating common random-effects formulations.
11. Custom Model Formula
Advanced users may override the automatically generated model equation and manually define complex models.
12. Multiple Model Analysis
FlexibleGLMM supports analysis of:
- Multiple dependent variables
- Multiple independent variables within the same workflow, reducing repetitive model specification.
Output Tabs
Instructions
Provides:
- Application instructions
- Usage examples
- Random-effects syntax guide
Data
Displays:
- Original or processed datasets
- Variable information
- Data tables
FitDist Output
Provides:
- Distribution comparison statistics
- Goodness-of-fit metrics
- Ranked candidate distributions
Fit Distribution Plots
Provides visual assessment through:
- Histograms
- Density curves
- Q-Q plots
Model Output
Displays:
- Model formula
- Fixed-effects estimates
- Random-effects estimates
- Standard errors
- Confidence intervals
- Model statistics
ANOVA
Provides:
- Type I/II/III tests
- F-statistics
- Degrees of freedom
- p-values with SAS-like Type III tests through afex.
Summary Table
Generates publication-ready tables summarizing:
- Model estimates
- Statistical significance
- Confidence intervals
Performance
Evaluates:
- Model fit
- R² statistics
- Information criteria
- Additional performance metrics
Post-hoc (EMMs)
Computes:
- Estimated marginal means (least-square means)
- Pairwise comparisons
- Contrasts
- Multiple comparison corrections using the emmeans package.
Summary Plots
Creates:
- Effect plots
- Estimated marginal mean plots
- Interaction plots
- Confidence interval visualizations
Diagnostics
Provides comprehensive model checking using:
- Residual plots
- Q-Q plots
- Simulated residual diagnostics
- Influence diagnostics
- Outlier assessment using packages such as DHARMa and influence.ME.
Auto Report
Automatically generates reproducible reports containing:
- Data preprocessing summary
- Model specification
- Statistical results
- Figures
- Diagnostic outputs
R Packages Used
User Interface
- shiny
- DT
- shinyjs
Data Handling
- readr
- readxl
- dplyr
Mixed Model Analysis
- afex
- lme4
- nlme
- emmeans
- predictmeans
Model Evaluation and Diagnostics
- performance
- DHARMa
- effectsize
- parameters
- r2glmm
- lookout
- fitdistrplus
Reporting and Visualization
- ggplot2
- broom.mixed
- knitr
- rmarkdown
- kableExtra
Parallel Computing
- future
Example Applications
FlexibleGLMM can be applied to:
- Longitudinal clinical studies
- Repeated-measures experiments
- Neuroimaging studies
- Behavioral experiments
- Multicenter clinical trials
- Educational and social science research
- Hierarchical and clustered datasets
Future Development
Planned enhancements include:
- Additional GLMM distributions
- More covariance structures
- Enhanced model comparison tools
- Improved visualization modules
- Extended reporting options
- Expanded support for SAS model replication
Citation
Puneet Talwar, Fermin Balda Aizpurua, Christophe Phillips, Gilles Vandewalle (2026, July 30). FlexibleGLMM: An R Shiny Application for Generalized Linear Mixed Models Across Multiple Statistical Engines: First formal release (Version 1.1.0). Zenodo. DOI: 10.5281/zenodo.21410757
Contact
Developer: Puneet Talwar University of Liège, Belgium
- For feedback/queries, bug reports, or feature requests, please send an email to ptalwar@uliege.be; talwar.puneet@gmail.com.