General Analysis Workflow

Most guided procedures follow the same pattern:

Confirm the dataset, filters, split files, weights, and missing-data settings.
Open an analysis from the Analysis menu.
Select dependent, independent, grouping, time, panel, or model variables as required by the procedure.
Configure options such as confidence levels, model terms, bootstrap settings, missing-data handling, diagnostics, plots, and output details.
Run the procedure.
Review warnings, assumptions, omitted cases, tables, charts, and model summaries in Results.
Export only the output needed for reporting, review, or archiving.

For confidential projects, confirm that selected variables, labels, output titles, charts, and exported filenames do not expose private identifiers before sharing results.

Choosing Variables

SCCalc uses variable metadata to guide setup, but you should still confirm every role. For example:

Grouping variables should be categorical and contain the intended groups.
Scale variables should be numeric and measured at an appropriate level.
Time-series variables should be ordered consistently.
Panel identifiers should uniquely identify entities.
Model terms should match the research design, not only the column names.
Weight and filter settings should be intentional and documented.

If a dialog does not offer the variable you expected, check Variable View for type, measurement level, and missing-value metadata.

Procedure Families

The Analysis menu groups procedures by use case. Use the family names to orient your choice, then rely on the procedure dialog, SCL Quick Reference, and Command Reference for detailed syntax and options.

Descriptive Statistics: summaries, frequencies, crosstabs, chi-square tests, Q-Q plots, and multiple response summaries.
Correlation & Regression: correlations, partial correlations, distance correlation, linear regression, linear models, curve estimation, and related model families.
Compare Means: t tests, ANOVA, GLM univariate workflows, and common nonparametric alternatives.
General Linear Model and Generalized Linear Models: univariate, multivariate, repeated measures, logistic, ordinal, probit, and multinomial workflows.
Mixed Models: linear mixed models, generalized linear mixed models, and generalized estimating equations.
Dimension Reduction and Scale: factor analysis, principal components, optimal scaling, proximity mapping, and reliability analysis.
Classify, Cluster, and Market Research: discriminant analysis, nearest neighbor, decision trees, neural networks, ROC curves, clustering, conjoint, and multidimensional unfolding.
Mediation & Moderation: mediation, moderation, parallel mediation, moderated mediation, and cross-lagged panel workflows.
Time Series, Panel Data, Forecasting, and Survival: ordered data, panel effects, stationarity and volatility workflows, forecasts, and survival models.
Bayesian, SEM, Utilities, Geospatial, and Meta-Analysis: specialized procedure groups where enabled.

This manual does not replace a statistical methods text. It explains how to use SCCalc safely and reproducibly. Choose methods based on your design, data collection process, assumptions, and reporting requirements.

Assumptions And Warnings

Treat warnings as part of the analysis result. Review:

Sample size and omitted cases.
Missingness and filtering.
Constant or near-constant variables.
Singular matrices and collinearity.
Convergence messages.
Model-fit diagnostics.
Distributional or residual assumptions.
Category counts and sparse cells.
Warnings about unsupported options or fallback behavior.

Do not copy a coefficient, p value, fit index, or chart into a report until you understand any warning that appears near it. If a result looks wrong, return to the data and metadata before changing model options.

Run Again And Workflow Builder

After a successful procedure, use Analysis -> Run Again to repeat the last analysis when the same setup is still appropriate. Use Analysis -> Analysis Workflow when you need a more structured multi-step workflow.

For work that must be repeated across datasets or over time, convert the stable steps into SCL. Manual dialog setup is useful for discovery; SCL is better for review, repetition, and audit trails.

Reproducibility

For every analysis you may need to defend or rerun, record:

Source data and import date.
Variable transformations and scale construction.
Filters, split files, and weights.
Procedure name and selected variables.
Missing-data handling.
Confidence levels, bootstrap settings, and model options.
Export format and output filename.
Any warnings that affect interpretation.

If an SCL script will be shared, review it for private file paths, dataset names, comments, embedded values, passwords, tokens, and organization-specific labels. Replace those details with neutral placeholders before sending the script outside your trusted team.

Running Analyses