Statistical Software by Analysis Type
Which Program Fits Which Method
Statistical software by analysis type means choosing the program from the planned method: one with a tested routine for that analysis and a record a committee or journal can check. No program leads on every method, so the plan comes first and the program follows from it.
Statistical software by analysis type is chosen from the planned method: the program that offers a well-tested routine for it, records every step so that the result can be repeated and produces files that a supervisor, committee or journal can open and verify. No program leads on every method, which is why the analysis plan comes before the choice of software. A group comparison, a multilevel model, a structural equation model, a survival analysis, a meta-analysis and a set of coded interviews each have different strong options. At OnlineSPSS, a specialist statistical consultancy for academic, clinical and professional research, an in-house PhD-level statistician carries out the agreed analysis in the program that suits the method, while the choice of software and the conclusions belong to the researcher.
In This Guide
BY ANALYSIS TYPE
Method Families and the Programs That Suit Each
The planned model narrows the field before anything else does, since a program that is strong for one family of methods can be thin for another.
Group Comparisons and Regression
IBM SPSS, R, Stata, SAS, jamovi and JASP all run these standard analyses, so the choice rests on output style, the record kept and who must open the files.
Tip: Run one familiar analysis in two candidate programs on a small dataset. Matching results build confidence, and any difference exposes a default worth knowing about.
Multilevel and Longitudinal Models
R, Stata, SAS, IBM SPSS and Mplus fit mixed and growth models, and they differ in the estimators, covariance structures and diagnostics that they offer.
Check: Confirm that the program offers the covariance structure and the estimation method your design needs. Defaults differ between programs, and the choice changes the standard errors.
Latent Variable and Structural Models
Mplus, R with the lavaan package, the sem commands of Stata and Amos fit structural equation models, and SmartPLS fits partial least squares path models.
Watch: Covariance-based and partial least squares models answer different questions. The research question settles the method first, and the program that implements it comes second.
Survival and Clinical Data
SAS, Stata, R and IBM SPSS fit Kaplan-Meier and Cox models, and SAS is often used where regulated clinical work calls for documented, validated programs.
Check: Ask whether the program handles the features of your data, such as time-varying covariates or competing events, before the choice is final.
Pooling Results Across Studies
R and Stata fit meta-analytic models, while RevMan and Comprehensive Meta-Analysis offer guided steps for standard syntheses of study-level results.
Tip: Choose a tool that reports heterogeneity statistics and exports forest plots in the format a journal requires, and note the version used in the review protocol.
Interview and Document Coding
NVivo, MAXQDA and ATLAS.ti organize coding, memos and retrieval for qualitative material, and the interpretation of the codes remains the researcher’s work.
Check: Test an export of the code list and the memos before committing, so that the coding framework can be shown to a committee or an examiner.
REVIEWER REQUESTS
What a Committee or Journal Asks of the Software
These requests recur whichever program is used, and the table shows how a menu-driven package and a script-based program usually answer each, without ranking either.
| Request | Menu-driven package | Script or command-based program | What to confirm |
|---|---|---|---|
| Show how a table was produced | Saved syntax, if each dialog was pasted | A script or command file that reruns from the first line | That the record reruns on the saved data |
| State the version | Release of the program and of each module | Release plus each package or contributed command | That versions are noted when the analysis is run |
| Share the files | Data and syntax files, opened by the same program | Plain-text script and data, readable in many tools | That recipients hold a compatible program |
| Rerun on another machine | Needs a licensed copy with matching modules | Needs the language or program and its packages | That access outlasts the project |
| Fit a newer method | Depends on the release and licensed modules | Often through a contributed package | That the routine reproduces a known result |
SWITCHING COSTS
When a Change of Program Pays Its Way
Staying put costs little, while a move in mid-project costs recoding, relearning and re-checking, so the points below weigh when that price is worth paying.
Freeze the Program at the Analysis Plan Statistical software by analysis type is settled when the plan names its models, and the choice holds unless a planned model proves unavailable, since every later change repeats data preparation and checking.
Reproduce One Result Before Moving Before a project changes program, one finished analysis is rerun in the new one and compared with the old output, so a gap in the routine is found while the original is still at hand.
Align Settings Before Comparing Where earlier tables exist, the new program's options are set to match the old ones first, so that a remaining difference is a finding about the data and not a leftover setting.
Record the Switch in the Methods A change made after the analysis began is stated with its reason, the version of both programs and the analyses affected, because a reader who meets two programs unexplained will ask why.
Compare the Criteria in Full Record, files, versions and access can be weighed against each other, and statistical software for dissertations is compared on exactly those criteria in Choosing Statistical Software for Research.
Switch Only for a Missing Model A change pays its way when a planned model is unavailable or collaborators need a different file format, and rarely for familiarity alone, since the recoding and the reproduction of earlier results fall on the same deadline.
COMMON ISSUES
Software Decisions Under a Reviewer's Eye
Nothing here triggers a warning in the program; each habit shows only when a second person tries to check or repeat a result.
Popularity Read as Quality A program's share of citations shows what a field uses, not whether it suits a particular model, so the plan is checked before an example from the literature is copied.
A Menu Entry Mistaken for Fit A dialog with the right name may not offer the variant a design needs, such as a given covariance structure or weighting scheme, so its options are compared with the plan.
Add-Ons Left Out of the Plan Some analyses need licensed modules or contributed packages, so a program counts as suitable only once those are confirmed to be available and are named.
Output Without Its Settings Tables pasted from a program without the options behind them cannot be checked, so syntax or a script is saved for every table that reaches the manuscript.
Walkthroughs Taken as Method Following a walkthrough built for a different design reproduces its choices and not the ones this study needs, so the model is checked against the plan before any guide is used.
Handover Overlooked A project that only one person can open stalls when a reviewer asks for the files or the analyst moves on, so a second person is able to run the record from start to end.
REPORTING
Naming the Program, the Routine and the Version
Two programs can give one model the same label and fit it differently, so the methods section records which routine produced each result.
For each model, the methods section names the program, the routine (the procedure, command, package or function) and the estimator or algorithm used. APA 7 gives specialized software a reference entry with developer, year, title and version, while widely used programs may need only their name and version in the text, so the handbook or journal settles which applies. For R, the citation function prints the reference text for R itself and, given a package name, for that package. Add-on modules and contributed commands are named separately. A data and code availability statement says what is shared, in which format and where, and a note that syntax or scripts are provided lets a committee judge whether the analysis can be rerun.
SOFTWARE
IBM SPSS, R, Stata, SAS, Mplus and NVivo in Detail
Each page shows what is analyzed in that program, what the work starts from and what is delivered with the results.
IBM SPSS
Menus with a syntax record
R
Scripts that rerun an analysis
Stata
Commands saved in do-files
SAS
Programs and logs for clinical data
Mplus
Latent variable and growth models
NVivo
Coding interviews and documents
ANALYSIS SERVICES
From the Analysis Plan to a Second Reading
The analysis plan, the analysis itself, the coding of qualitative material and a second reading each lead to different support, whichever program holds the data.
Research Planning & Study Design
The analysis plan that names the models, so the program is chosen from the plan and not before it.
Quantitative Data Analysis
Models estimated in the program that implements them, handed over with the syntax or script that produced each table.
Qualitative & Mixed Methods Analysis
Interview and document coding organized into a framework, in NVivo, MAXQDA, ATLAS.ti or the program you already use.
Independent Statistical Review
A second look at an analysis already run, from its settings and defaults to the tables it produced.
FREQUENTLY ASKED QUESTIONS
Second Programs, Disagreeing Results and Committee Rules
A model that one program lacks, two programs that disagree and a committee with a program of its own each call for a different kind of answer.
What if one planned model needs a second program?
Two programs can serve one project, provided the link between them is documented. Some projects fit most models in one program and a specialist model, such as a latent variable model, in another. Clean the data once, export from that file for the second program, and name the version of each program in the methods section.
Why do two programs give slightly different answers for one model?
Defaults differ: the estimation method, the treatment of missing values, the degrees-of-freedom rule and the optimizer can all vary. Compare the options and not the labels, align them where possible and report the settings used. Small numerical differences from an optimizer are expected, while large ones suggest that two different models were fitted.
Can a committee require a particular program?
Some handbooks name one, and the requirement usually concerns the files the committee is able to open. Where the named program lacks a planned model, the methods state which program fits it and why, and agreeing that with the supervisor early avoids a dispute at the defense.
Must the program be chosen before I contact OnlineSPSS?
No. The program is settled during the assessment from the planned models, the files you hold and any rule your degree program or journal sets. If a planned model needs a program you do not hold, that is flagged when the scope is confirmed, and the final choice remains yours.
How do you protect my research data and confidentiality?
Your dissertation materials and project details are handled confidentially. Project information is not published, and 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 a protocol, an analysis plan or a dataset, 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
Which Program Fits
Your Planned Analyses?
Three facts narrow the options: the models the plan calls for, the program a supervisor or journal expects and the files already in hand. An assessment weighs them together.
No Commitment Required
Strictly Confidential
Fixed Project Fee
In-House PhD-Level Statisticians
Same-Day Response
Serving Researchers Since 2014