The Outcome-Based Wizard
The Outcome-Based Sample Size Wizard is the guided engine that the "I Know My Goal" and "I Know My Outcome" pathways funnel into. It walks you from your outcome type, through your study design, to a configured calculation with clinically-grounded effect-size guidance at each step.
The three steps
outcome --> design --> calculator
| Step | What you do |
|---|---|
| Outcome | Choose the type of outcome you will measure (skipped if already selected from a pathway) |
| Design | Choose your study design (independent groups, paired/matched, case-control, cohort, etc.) |
| Calculator | Review recommended configuration, adjust parameters, get sample size or power |
Step 1 -- Outcome
The outcome determines the entire downstream calculation. Outcomes fall into categories (continuous, binary, correlation, survival, diagnostic, etc.), each with built-in effect-size guidance -- small/medium/large reference values drawn from the literature.
Examples:
- Continuous clinical outcomes (lab results, QoL scales, walk tests) bring their own effect-size anchors
- Binary outcomes (mortality, adverse events) configure proportion-based calculations
- Odds ratio or relative risk configure epidemiological calculations
Some outcomes also set sensible defaults (e.g. a default control-group proportion).
Step 2 -- Study design
Design options span the full range the module supports:
- Two or More Independent Groups -- parallel-group comparison
- Paired/Matched Groups -- within-subject or matched design
- Multiple Treatment Groups -- dose-response or multiple arms
- Case-Control Study -- for odds-ratio outcomes
- Cohort Study -- for relative-risk outcomes
- Compare Two Correlations / Compare Two Diagnostic Tests
- Agreement/reliability variants (ICC, Bland-Altman, categorical agreement)
- Bioequivalence (crossover and non-crossover)
- Factor analysis, Mann-Whitney U
The design + outcome combination determines which calculator the wizard routes to:
| Outcome + design | Routes to |
|---|---|
| Continuous + Independent Groups | Independent means calculator |
| Continuous + Paired/Matched | Paired means calculator |
| Binary + Independent Groups | Independent proportions calculator |
| Odds-ratio + Case-Control | Odds-ratio sample size calculator |
| Relative-risk + Cohort | Relative-risk sample size calculator |
Step 3 -- Calculator and recommendations
The wizard presents the configured calculation with power recommendations -- typically showing sample size at 80% and 90% power for small/medium/large effects.
Adjustable parameters (depending on calculation):
- Effect size (or components: mean difference, SD, proportions, OR/RR)
- Correlation (for paired designs)
- Control proportion (for binary outcomes)
- Margin of error (for prevalence/precision designs)
- Number of groups (for multi-group designs)
- Coefficient of variation
- Prevalence and population size (for survey designs)
Built-in effect-size guidance
The wizard includes an assistant panel providing clinically-grounded, literature-based guidance for translating your study into an effect size. Worked examples for each outcome:
- HbA1c: "If mean reduction = 0.5% and SD = 1.0% then Cohen's d = 0.50. For paired design: pre-post correlation rho = 0.50-0.70 reduces sample size significantly."
- eGFR: "If mean difference = 8 mL/min and SD = 20 then Cohen's d = 0.40. For paired: correlation rho = 0.60-0.80 (reduces sample to ~40-60% of independent design)."
The assistant provides pre-prepared domain-specific guidance and worked examples. The actual computations are performed by the calculators' validated formulas -- the assistant guides parameter choice, not statistical calculation.
When to use the wizard vs direct calculators
| Use the wizard when... | Use a direct calculator when... |
|---|---|
| You know outcome and design but not the exact test | You already know which calculator you need |
| You want literature-based effect-size guidance | You have your effect size ready |
| You want to see sample sizes across effect-size tiers | You want one specific computation |
Both routes use the same underlying formulas and produce the same Methods Section output.