Missingness, Multiplicity, and Degrees of Freedom

Use this section when missing data, multiple comparisons, or repeated analytic choices could change how strongly a result should be interpreted.

When to use this category

Use this category when your question is about whether missing data, multiple comparisons, or repeated analytic choices change how strongly a result can be interpreted.

This section covers missing-value handling, analysis-family multiplicity, and researcher degrees of freedom; it is not the place for study design and independence questions or for model-diagnostics questions.

If your main question is whether observations are independent, whether batches or repeated measures change the design, or whether a regression model is mis-specified, go to Design and Independence or Model Diagnostics instead.

Quick routing

If your main question is...Go to...Why
Are missing values, imputation, or dropped rows changing the analysis N or what must be disclosed?Missing Data and AttritionCovers missingness detection, imputation options, attrition, and MCAR / MAR / MNAR assumptions
Are there multiple pairwise comparisons in one figure, or several claim-bearing figures addressing the same question?Multiplicity and Analysis FamiliesCovers within-figure multiplicity, analysis families, and family-wide policies such as Holm correction
Are outlier exclusions, subgroup choices, or retesting steps making the analysis more flexible after seeing the data?Outliers and Researcher Degrees of FreedomCovers outlier removal, exploratory branching, alpha-inflation heuristics, and disclosure of repeated analytic choices

Child pages

Start here

  1. Start with Missing Data and Attrition if the main issue is missing values, imputation, dropped rows, or attrition disclosure.
  2. Start with Multiplicity and Analysis Families if the main issue is Tukey HSD, Holm, Bonferroni, pairwise comparisons, or several related claim-bearing figures.
  3. Start with Outliers and Researcher Degrees of Freedom if the main issue is outlier exclusion, subgroup selection, retesting, or p-hacking risk.
  • Related links are registered in frontmatter and rendered below this stub.