Assumptions and Robustness

Use this section when you need to check whether assumptions, outliers, or reasonable analysis variations could change how you interpret a result.

Use this section when your question is whether a result depends too heavily on distributional assumptions, variance assumptions, outliers, or reasonable analysis alternatives, and you need to choose the right robustness-oriented page to read next.

This section is for checking how fragile or stable an analysis may be once you broadly know the analysis path. It is not the right place to choose the primary method itself or to assume that passing these checks alone proves a result is valid; if your first question is which test or model to run, start with Methods, and if your first question is about dependence between observations or design structure, start with Design and Independence.

Is this the right section?

Use the table below to choose the leaf page that best matches the assumption or robustness question you need to answer.

If your question is...Start here
Are normality or equal-variance assumptions plausible enough for a standard parametric path?Normality and Homoscedasticity Checks
Do I need a high-level warning about fragile assumptions and what should be disclosed?Assumption and Robustness Guard
Would the conclusion change under reasonable alternative specifications or analysis choices?Sensitivity Analysis Engine
How does the direction or strength of the conclusion change across alternative choices?Conclusion Sensitivity Profile
Do many defensible analytic paths lead to different answers?Multiverse Analysis
Are the conclusions materially affected by extreme or influential observations?Outlier Sensitivity Report

If your main task is choosing the core comparison or model family, this is usually not your first stop. Use Methods before returning here.

Support boundary

This category covers assumption checks and conclusion sensitivity, not a standalone guarantee that an analysis is correct, complete, or automatically robust.

Child pages

Start here

  1. Start with Normality and Homoscedasticity Checks if you first need to judge whether common parametric assumptions look plausible.
  2. Continue to Assumption and Robustness Guard if you need a conservative summary of assumption-related risk and disclosure needs.
  3. Go to Sensitivity Analysis Engine if the main question is whether reasonable alternative specifications change the result.
  4. Use Conclusion Sensitivity Profile when you need to summarize how conclusions shift across alternatives.
  5. Use Multiverse Analysis when multiple defensible analytic paths may lead to meaningfully different answers.
  6. Use Outlier Sensitivity Report when influential observations may be driving the finding.
  • Related links are registered in frontmatter and rendered below this stub.