Experimental Design

Routing Summary

This folder covers tools for designing, powering, and analyzing experiments. Contains 3 notes.

Concept Map

ConceptNoteTypeDepends OnKey Result
Sample size formulas, effect sizes, practical guidelinesPower Analysis and Sample SizeconceptThe Experimental Ideal, Garden of Forking Paths, Researcher Degrees of FreedomPower = P(reject H0 given H1 true); aim for 80%+
Bonferroni (FWER), Benjamini-Hochberg (FDR), q-valuesMultiple Testing CorrectionsconceptGarden of Forking Paths, Researcher Degrees of Freedom, Power Analysis and Sample SizeFDR control is usually more appropriate than FWER
Type S (sign) and Type M (magnitude) errorsType S and Type M ErrorsconceptMultiple Testing Corrections, Power Analysis and Sample Size, Multiple Comparisons - Bayesian PerspectiveSign and magnitude errors matter more than Type 1 in social science
Kaplan-Meier, log-rank test, Cox proportional hazardsSurvival AnalysisoverviewThe Experimental Ideal, Regression and the CEF, Power Analysis and Sample SizeCensoring requires specialized time-to-event methods

Notes

  • Power Analysis and Sample Size — CONTAINS: Sample size formulas, effect size conventions, power curves, practical guidelines for experiments
  • Multiple Testing Corrections — CONTAINS: Bonferroni correction (FWER), Benjamini-Hochberg (FDR), q-values, when to use each method
  • Type S and Type M Errors — CONTAINS: Sign errors, magnitude errors, exaggeration ratio, why large estimates from small samples are misleading, connection to underpowered studies
  • Survival Analysis — CONTAINS: Kaplan-Meier estimator, log-rank test, Cox proportional hazards regression, censoring mechanisms

Sources

See Also