Frequently asked questions.

Surv.lytics is an integrated workspace for preparing survey datasets, running supported quantitative and qualitative analyses, and organising reproducible findings.

It is designed for students, academics, evaluators, and applied researchers working with survey data. Beginner guidance is available, while methodological decisions remain the researcher's responsibility.

You can begin with Guided Research and plain-language help. You should still understand your research design and review assumptions before interpreting any result.

It builds an evidence-based sequence from the current project and dataset, explains prerequisites, and suggests next steps. Suggestions never confirm methodology automatically.

The current importer supports CSV, Excel XLSX, and SPSS SAV files within configured size, row, and column limits.

Yes. Upload a .sav file through the authenticated dataset upload workflow. Labels and supported metadata are profiled during ingestion.

It profiles storage types, measurement-level candidates, missingness, and other metadata. These are reviewable suggestions rather than final research classifications.

No. The source upload is stored as an immutable version. Approved transformations create a separate working version with an audit trail.

The workspace can organise researcher-approved roles and provide contextual suggestions. It does not infer causal roles from column names or values alone.

Missing counts and patterns are presented for review. Deletion, imputation, or other changes require a separate preview and explicit approval.

Yes. Reliability analysis is available for appropriate multi-item sets, with item diagnostics and version-bound provenance. Reliability is not treated as proof of validity.

Previewing or opening an analysis does not create final evidence. Completed runs and explicit save actions preserve configuration and dataset-version provenance.

Yes. Immutable completed analyses remain available in history and can be organised in Results Notebooks, including when newer specifications exist.

It calculates covariance-based coefficient alpha from the selected scored items and reports item diagnostics. Listwise mode uses one complete-case covariance matrix; pairwise mode reports its smaller pair counts and limitations. The technical Reliability article documents the equation and implementation.

They are not interchangeable labels. Alpha is covariance-based internal consistency; omega uses a congeneric factor model; composite reliability uses fitted CFA or PLS measurement quantities; rho_A is reported only in its implemented PLS context. Review each method's assumptions before comparing values.

Yes. It supports the documented continuous-indicator CFA and recursive CB-SEM workflows, plus supported recursive PLS-SEM models. The Methods directory identifies workflow-only capabilities.

CB-SEM estimates a covariance-based latent model with global fit evidence. PLS-SEM uses a component-based path-modelling workflow with separate measurement and structural assessment. Method choice requires theoretical and design justification.

The supported continuous-indicator CFA and recursive CB-SEM implementations have frozen numerical benchmark fixtures against documented lavaan versions and declared tolerances. The Knowledge Base also discloses library versions, estimator boundaries, known differences, and unsupported options.

No blanket equivalence is claimed. Surv.lytics implements and tests a documented path-weighting PLS-PM engine, with a frozen seminr comparison for supported statistics. Its repeated K-fold prediction assessment is explicitly not labelled PLSpredict.

Yes. You can detect candidate text fields locally, confirm questions, build a versioned codebook, assign codes, approve themes, and create mixed-methods displays.

No. Statistical engines produce numerical evidence. AI assistance, when configured, provides separate explanatory or drafting support that must be reviewed.

Respondent data is not sent by default. The assistant uses bounded context and shows the configured privacy boundary; researchers should avoid adding identifying content to prompts.

Supported reporting workspaces can generate DOCX, PDF, HTML, CSV, or XLSX outputs depending on the artefact. Review every export before sharing it.

Application queries scope projects, datasets, analyses, and reports to the authenticated owner. Public help and marketing pages contain no project or respondent data.

Yes. Use My Profile in the account menu. Password changes use Django's validated password workflow and retain the authenticated session.