STATISTICAL SOFTWARE
Mplus Statistical Analysis Service
Latent Variable Models You Can Audit
Our Mplus statistical analysis service carries out latent variable, mixture, growth and multilevel models in Mplus, working from your data and .inp files, with the input and output files of every model version delivered where agreed so that each revision stays traceable and rerunnable.
No Commitment Required
Strictly Confidential
In-House PhD-Level Statisticians
Fixed Project Fee
Same-Day Response
What Is Mplus Used For in Research?
Mplus is a commercial latent variable modeling program from Muthén & Muthén, used for confirmatory factor, structural equation, growth, mixture and multilevel models, including models with categorical outcomes and missing data. Researchers choose it for estimators matched to the scale of the indicators, for latent class and growth mixture models, and because the input file states the whole model in text. The .inp file shows what was asked and the .out file shows what happened, so every model can be audited. OnlineSPSS, a specialist statistical consultancy for academic, clinical and professional research, provides an Mplus statistical analysis service in which an in-house PhD-level statistician carries out the agreed models in Mplus, working from your files and keeping each model version.
WHEN TO ENGAGE
Four Moments When Researchers Bring Us an Mplus Model
A project reaches us with a model to estimate, a mixture solution that will not settle, syntax that someone else wrote or an extension across groups or levels.
Model Specified, Estimator Open The measurement or structural model is decided, and it has to be estimated with an estimator that suits categorical or continuous indicators and your missing-data structure, then reported with fit and parameters.
Mixture Solution That Will Not Settle A latent class or growth mixture solution changes between runs, a class is very small or the best log-likelihood is not replicated, and the number of classes is in doubt.
Syntax to Audit or Extend A coauthor's .inp files or a chain of model revisions need checking, and an Mplus statistician reads the syntax and output in full, not only the final table of estimates.
Groups, Time or Clusters Added A scale has to be compared across groups or waves, or the data are clustered, and the model must be extended in a documented sequence of steps that can be followed later.
ANALYSES IN MPLUS
Analyses Carried Out in Mplus for Your Latent Variables
Mplus SEM analysis and its mixture and multilevel relatives start from the scale of your indicators and the structure of your data, and each card links to the method page behind the choice.
Structural Equation Modeling (SEM)
Measurement and structural parts are specified together, with the estimator matched to the indicators and direct, indirect and total effects reported with intervals.
Confirmatory Factor Analysis (CFA)
Items declared as categorical receive a categorical-data estimator, loadings and thresholds are read together, and fit is judged from several indices with the theory in view.
Latent Class Analysis (LCA)
Mplus latent class analysis enumerates solutions from random starts, compares them by information criteria and interpretability, and carries classification uncertainty into later analysis.
Growth Mixture Modeling (GMM)
Trajectory classes are fitted with their variances specified step by step, and the solution is checked for replication, class size and substantive meaning.
Multilevel SEM (MSEM)
Within-level and between-level models are specified separately, intraclass correlations are examined first, and the number of clusters is weighed against model complexity.
Measurement Invariance
Configural, loading and intercept or threshold steps are tested in sequence, using the steps that suit the indicator type, before group or wave means are compared.
COMMON MPLUS ISSUES
Where Mplus Models Go Wrong Before the Fit Table
An Mplus model can run and print estimates while the setup is wrong, so each check below is made before a result goes into a table.
Estimator and Indicator Scale Ordinal items with few categories that are not declared as categorical are fitted as continuous, which can bias loadings and fit. The categorical declaration and the estimator are matched to every indicator before the model is run.
Missing-Data Handling Maximum likelihood and weighted least squares estimators treat incomplete data differently, so the choice is checked against the pattern of missingness and the assumption it requires.
Random Starts and Replicated Likelihood A mixture solution is accepted only when its best log-likelihood is replicated from many random starts, because one converged run can sit on a local maximum.
Warnings in the Output A model can terminate normally and still carry a warning about a non-positive definite matrix, a negative variance or a saddle point. The output is read line by line, not only its fit summary.
Variable Order and Missing Codes The data file has no header, so the order of the variable names and the declared missing-value code give each column its meaning. A misaligned list produces plausible but wrong results.
Nested Model Comparisons With the WLSMV or MLR estimators, the chi-square values of two nested models cannot simply be subtracted, so comparisons use the DIFFTEST option or the scaling correction factor, as the estimator requires.
DELIVERABLES
What You Receive From an Mplus Analysis
The files follow a traceable Mplus project, agreed in the scope and provided where applicable, so that every revision of the model can be found and rerun.
Input and Output Files The .inp and .out files of each model reported, delivered as run, so that every table traces to the syntax and output that produced it.
Model Version Log A log of the model sequence, from the first specification to the final one, with what changed at each step and why.
Prepared Data File The data file as Mplus reads it, with a key to the variable order and the missing-value code, so that columns cannot be misread later.
Fit and Parameter Tables Publication-ready tables of fit indices, estimates and intervals, with the sample size and the estimator stated beside every model.
Statistical Interpretation Each parameter and each latent class is explained in the terms of your hypotheses, including what the model cannot establish, such as causal order in cross-sectional data.
Software Version Record The Mplus version and the main estimation settings are recorded, so that the project team can rerun the models later in the same setup.
OTHER PLATFORMS
When Another Platform May Be More Appropriate
Mplus is a strong fit for categorical, mixture and multilevel latent models, and a project with simpler needs or other collaborators may fit another platform better.
ILLUSTRATIVE EXAMPLE
How an Mplus Analysis May Look in Practice
An illustration of how a latent class model built on yes or no items moves from an unreplicated solution to one that can be audited.
ILLUSTRATIVE EXAMPLE · ADOLESCENT RISK BEHAVIOR RESEARCH · UNITED STATES
A research team studying adolescents wanted to know whether young people fall into distinct groups defined by six yes or no risk behavior items collected in a school-based survey.
The items were declared as categorical, so the model used an estimator suited to binary indicators. A first run with a small number of random starts returned a three-class solution whose best log-likelihood appeared only once, which was a sign of a possible local maximum. The number of random starts was increased until the best value was replicated, solutions with two to five classes were compared by information criteria, entropy and interpretability, and a very small class was examined before it was accepted. Every model in the sequence was kept as its own input and output file, so that the path to the final solution could be followed.
Outcome: The team received the class solution with its fit statistics, the item response probabilities with their intervals, the full set of .inp and .out files and a log of the model sequence.
This example does not describe a real client or project. We never share details of actual client work, in line with the confidentiality we promise every researcher. This illustrative example is included to show the type, scope, and standard of work involved in this kind of analysis.
RESEARCH DISCIPLINES
Research Fields Where Mplus Is Common
Mplus is often used in psychology, education and the health and social sciences for measurement, growth and person-centered models, and the same estimator questions arise wherever latent constructs are measured.
HOW WE ASSESS YOUR PROJECT
We Review Your Actual Research
Your assessment is based on a review of your actual research materials and analytical requirements, not a generic project category.
You Submit Your Project Details
Share whatever is available at the current stage, such as the research aim, protocol or proposal, dataset, variable list, manuscript or statistical requirements.
We Review Your Research and Define the Scope
Your statistically trained Research Coordinator reviews the materials to understand the study design, research questions, data structure, analytical requirements and expected deliverables. If clarification is needed, additional documents or information can be requested before the assessment is finalized.
Your Analysis Is Completed and Delivered
Once the scope and fixed project fee are agreed, the in-house statistician completes the work and the Research Coordinator remains the communication point through delivery, with the analysis, deliverables and timeline as agreed.
RESEARCHER VOICE
What Researchers Say
Feedback from a psychometrician on matching the estimator to categorical indicators and an analysis the whole project team could rerun.
“Keeping the analysis in Mplus mattered because the rest of the project team needed to rerun it later. They dealt with matching the estimator to categorical indicators and the missing-data structure, documented the analytical steps clearly, and gave us outputs that were much easier to audit and report.”
Statistical Support Within Your Research OnlineSPSS provides the agreed methodological, analytical and statistical reporting support, including the Mplus files described in the scope. The researcher or research team retains the scientific question, substantive interpretation, conclusions and overall responsibility for the research.
FREQUENTLY ASKED QUESTIONS
Questions About Mplus Analysis for Research
Short answers to what researchers ask before sharing an Mplus project, from estimators and licenses to existing syntax, confidentiality and the assessment.
Which models do you estimate in Mplus?
At OnlineSPSS we estimate the models researchers most often run in Mplus: confirmatory factor analysis, structural equation models, latent class and growth mixture models, Mplus multilevel modeling, measurement invariance and models with categorical outcomes. The choice follows your research question and the scale of your indicators, and it is confirmed during the assessment.
How do you decide which estimator to use?
From the scale of the indicators, the missing-data structure, the sample size and the model, and the decision is documented in the delivered files. Categorical indicators and continuous ones generally call for different estimators, so the choice is checked at the start, not after a result looks odd.
Do I need an Mplus license to rerun or check the models?
To rerun them, yes: the .inp files are run in Mplus by whoever on your team holds a license. To check them, no: input and output files are plain text, so every setting, warning and estimate can be read and audited without the program.
Will my earlier model versions be kept when the model is extended?
Yes. Existing input and output files are read in full before anything is added, and earlier versions are preserved, so that each revision remains visible. A completed analysis can also be checked through Independent Statistical Review.
How do you protect my research data and confidentiality?
Your data files, input files and output are handled confidentially, and project information is not published. An NDA is available where required and can be arranged before sensitive materials are shared. Read more about confidentiality and data security.
What happens when I request an assessment, and how long does it take?
You share your research question, design and relevant materials, such as the dataset and any existing syntax, and a statistically trained Research Coordinator reviews the request and responds the same day. The analysis, deliverables, timeline and one fixed project fee are then confirmed before any work begins.
Still have a question? Find more answers in our full FAQ.
NEXT STEP
Discuss Your
Mplus Analysis With Us
Share the research question, the indicators and any input files you have, and the models you plan. We confirm the estimator and the data setup before any work begins.
No Commitment Required
Strictly Confidential
Fixed Project Fee
In-House PhD-Level Statisticians
Same-Day Response
Serving Researchers Since 2014