RESEARCH SUPPORT
Choosing the Right SPSS Procedure
GLM, MIXED and GENLIN for Your Design
In practice, choosing SPSS procedures means deciding whether your outcome, its clustering and its repeated measurements call for GLM, MIXED or GENLIN, with the reasoning documented beside the output for any reviewer or committee member to see.
No Commitment Required
Strictly Confidential
In-House PhD-Level Statisticians
Fixed Project Fee
Same-Day Response
How Are SPSS Procedures Chosen for a Given Design?
Choosing SPSS procedures begins with the structure of the data, because procedures that look alike in the menus answer different questions. The GLM procedures fit linear models for a continuous outcome and treat repeated measurements through a within-subjects factor, MIXED fits linear models with random effects and covariance structures for clustered or repeated data, and GENLIN fits generalized linear models for binary, count and skewed outcomes and, through its repeated option, marginal models by GEE. OnlineSPSS is a specialist statistical consultancy for academic, clinical and professional research; its in-house PhD-level statisticians make this choice from your design, fit the model in SPSS and record why the alternatives were set aside. Which question the model must answer remains your decision.
WHEN TO ENGAGE
Design Features That Make the Menu Choice Hard
Two procedures can run on the same file and return different answers, so the choice is normally settled before the analysis begins.
Several Measurements per Person Participants were measured at several times and some visits are missing, so it is unclear whether a repeated-measures GLM, a mixed model or a marginal model suits the data.
Participants in Clusters Pupils sit in classrooms or patients in clinics, so observations are not independent, and the procedure has to reflect that grouping instead of ignoring it.
Outcome Not Continuous The outcome is a yes-or-no event, a count or a skewed positive measure, and a linear model on the raw scale is a poor description of it.
A Reviewer Questions the Choice A journal reviewer asks why one procedure was used rather than another, and the justification must come from the structure of the data.
MATCHING DESIGN TO MODEL
How the Procedure Follows the Data Structure
Each family answers a particular question about the outcome and about the correlation among observations, and the choice among them is documented.
Outcome Scale Sets the Family
The scale of the outcome sets the model family: linear for a continuous measure, logistic for a binary event, Poisson or negative binomial for counts, within Regression & Predictive Modeling.
GLM or MIXED
The SPSS GLM vs MIXED question for repeated measures turns on missing visits: GLM Repeated Measures analyzes only participants with every measurement, whereas MIXED uses every observed value.
Random Intercepts and Slopes
Clustered data call for random intercepts, and sometimes random slopes, in MIXED, with the principles set out under Linear Mixed Models (Multilevel Modeling).
GENLIN for Other Outcomes
SPSS GENLIN models accept binomial, Poisson, negative binomial and gamma distributions with matching links, and overdispersion in counts is examined before a distribution is settled.
Marginal or Cluster-Specific
GEE gives an average effect across the sample under a working correlation, whereas a mixed model gives effects conditional on each cluster, as covered under Generalized Estimating Equations (GEE).
Documenting the Choice
The reason for the procedure, the options chosen and an alternative fitted as a check are recorded, so that the choice can be defended in a methods section.
WRONG TOOL, CLEAN RUN
Procedure Mismatches That Change the Answer
A procedure that runs without an error message can still be the wrong one, and the usual mismatches are easy to name once the structure of the data is set down.
Wide File, Complete Cases Only A repeated-measures GLM excludes every participant with one missing time point, so a study with scattered missing visits can lose a large part of its sample.
Clusters Treated as Independent Pupils or patients are analyzed with a procedure that assumes independent observations, so standard errors come out too small and the intervals too narrow.
Counts Fitted as Continuous A count with many zeros is analyzed as if it were continuous, which allows predictions below zero and standard errors that do not match the data.
GEE With Few Clusters A GEE model is fitted to a study with only a handful of clusters, although its robust standard errors need a reasonable number of clusters to be trusted.
Covariance Structure Accepted As Is A mixed model keeps its first covariance structure and no alternative is compared, although that choice affects the standard errors of the fixed effects.
Marginal Effect Read as Individual Change In a logistic model, a GEE odds ratio is read as change for an individual, though it is a population-averaged effect, usually closer to the null than the cluster-specific one.
DELIVERABLES
Documents That Show Why a Procedure Was Used
Alongside the fitted models, the record states the data structure and compares the alternatives considered, so that a methods section can cite its reasons.
Data Structure Summary A description of the unit of analysis, the clusters and the repeated measurements in the file, which is the basis for every later choice.
Procedure Comparison Note A short comparison of the procedures that could fit the design and why one was chosen, including what a different choice would have changed.
Commands for Each Model The SPSS commands for every model, with the distribution, link and covariance options stated, so that each model can be refitted from the file.
Fitted Estimates With Intervals Fixed-effect estimates, rate ratios or odds ratios with 95% confidence intervals and the number of observations and clusters used.
Alternative Specification Output Output from one alternative, such as a different covariance structure or distribution, showing how far the conclusions depend on the first choice.
Procedure Note for Methods A note on the model, its options and the software version, matched to the output, for you to check and incorporate.
ILLUSTRATIVE EXAMPLE
Repeated Counts Across Garden Plots
An illustration of how a count outcome observed repeatedly led to a different procedure from the one first considered.
ILLUSTRATIVE EXAMPLE · POLLINATOR GARDEN MONITORING RESEARCH · UNITED STATES
An ecology researcher counted bee visits to flowering plants in garden plots on repeated observation days and wanted to compare plots with a native wildflower strip against plots without one.
The researcher first planned a repeated-measures GLM on the daily counts. The complication was that the counts were small, with many zeros and a variance larger than the mean, and some observation days were missed because of weather. A GLM on raw counts would have excluded every plot with a missed day and assumed a continuous outcome. The procedure chosen was GENLIN with a negative binomial distribution and a log link, fitted as a marginal model with plot as the subject and an exchangeable working correlation. Because the number of plots was modest, a mixed model with a random intercept for plot was fitted as a check, and the report stated which result it relied on. Plots had not been allocated at random, so the findings were described as associations.
Outcome: The researcher received the rate ratio for strip versus no strip with its 95% confidence interval, the model comparison note, the SPSS commands and output, and a note on the procedure.
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.
SOFTWARE
Modeling Options Outside SPSS
SPSS reaches the families above through its GLM, MIXED and GENLIN procedures, and some designs are better served by a second program.
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.
RESEARCH RESPONSIBILITY
Modeling Support and Research Ownership
A documented procedure choice gives the researcher a basis for the methods section, and what the estimates mean for the field is for the researcher to judge.
Agreed Statistical Support OnlineSPSS provides the agreed methodological, analytical and statistical reporting support, defined in a scope that is confirmed before any work begins.
No Intellectual Contribution OnlineSPSS does not claim any intellectual contribution to your methods section, paper or report. Authorship remains the researcher's.
Your Research Decisions The research question, the substantive meaning of the estimates and the conclusions remain with you, along with overall responsibility for the study.
Confidential Handling Research data and project details are handled confidentially, project information is not published, and an NDA is available where required.
FREQUENTLY ASKED QUESTIONS
GLM, MIXED or GENLIN: What Researchers Ask
Three settings decide more SPSS model choices than the menus suggest: the GENLIN dispersion parameter, the line between GENLIN and generalized linear mixed models, and Random versus Repeated in MIXED.
What goes into choosing SPSS procedures for a study?
Choosing SPSS procedures at OnlineSPSS rests on three features of the data: the scale of the outcome, the clustering of participants and the repeated measurements, and each choice is documented together with the alternatives set aside. Where two models are both defensible, both can be fitted and compared.
Does GENLIN estimate the negative binomial dispersion parameter?
Only when asked to. In recent versions GENLIN can either fix the negative binomial ancillary parameter at a value, with 1 as the default, or estimate it from the data. A fixed value of 1 may not match the dispersion in the data, so estimating it is usually preferred for overdispersed counts, and the setting chosen is stated in the methods.
How does GENLIN differ from the generalized linear mixed models procedure?
GENLIN fits generalized linear models and, through its repeated option, GEE models that describe averages across the sample. The Generalized Linear Mixed Models procedure, found under Mixed Models in the editions and versions that include it, adds random effects, so its coefficients are conditional on the cluster. Both use distributions and links, and the choice follows whether the question concerns sample averages or effects within a given cluster.
What is the difference between the Random and Repeated settings in MIXED?
Random effects add variation between subjects, such as a random intercept, to the model. The Repeated setting instead specifies how the residuals of one subject correlate across occasions. Many models need only one of them, and using both can leave parameters poorly identified. The choice follows the design, and covariance structures that are comparable can be compared by information criteria such as AIC or BIC.
How do you protect my research data and confidentiality?
Your data and project details are handled confidentially, and project information is not published. Researcher names and institutions are not used publicly without permission. 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, a description of how participants were measured and grouped, and any output you already have, 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, so the timeline follows from the agreed scope.
Still have a question? Find more answers in our full FAQ.
NEXT STEP
A Procedure You Can
Defend to a Reviewer
GLM, MIXED or GENLIN is decided by how the outcome was measured and how participants are grouped, and a reviewer's question about the choice can be answered from those two facts.
No Commitment Required
Strictly Confidential
Fixed Project Fee
In-House PhD-Level Statisticians
Same-Day Response
Serving Researchers Since 2014