If you take only one rule from how to justify sample size research papers depend on, take this: every number needs a method behind it.
Sample size is the most-flagged element in quantitative methodology review, and the most-misunderstood element in qualitative review. Authors lose Round 1 reviews not because their sample was wrong, but because they never told reviewers why the sample was right. The fix is rarely more data — it's better justification.
This guide walks through the four sample-size justification approaches reviewers accept, the methods that map to each, and what a defensible paragraph actually looks like. For the broader methodology context, return to our pillar on how to write a methodology section that reviewers respect.
The four justification approaches that work
Sample-size justification comes in four flavours. Each is appropriate for different designs. The mistake most authors make is using the wrong approach for their method — defending a regression-based study with saturation language, or a phenomenological study with power analysis. Match the approach to the design.
A priori power analysis
For inferential quantitative work. State effect size, alpha, power, and the resulting minimum N. G*Power and pwr (R) are the standard tools cited.
Established ratios & rules of thumb
For SEM, CFA, multiple regression, factor analysis. Cite recognised parameter-to-observation ratios or published guidelines.
Saturation & information power
For qualitative work — interviews, focus groups, ethnographic studies. Defend the indicator and the point at which it was met.
Design-driven justification
For case studies, comparative designs, longitudinal panels. The logic of the design — not a calculation — defends the number.
Sample size minimums by method
The table below summarises typical floors. Treat them as starting points to defend, not as ceilings to assert. For survey-driven studies in particular, see our companion guide on quantitative research methodology — what to report and how.
| Method | Minimum N | Justification approach | Common reference |
|---|---|---|---|
| SEM | 200 | Parameter ratio (5–10 per estimate) | Kline (2016) |
| CFA | 150 | Factor loading × items | Hair et al. (2019) |
| Multiple regression | 50 + 8k | k = predictors; power analysis | Tabachnick & Fidell |
| EFA | 5–10 per item | Item ratio | Costello & Osborne |
| Mediation (bootstrap) | 200+ | Power analysis with bias-corrected CI | Hayes (PROCESS macro) |
| Thematic analysis | 15–30 interviews | Saturation / information power | Braun & Clarke; Malterud |
| Phenomenology (IPA) | 6–10 participants | Information power, deep cases | Smith & Osborn |
| Multi-case qualitative | 3–5 contrasting cases | Replication logic | Yin (case study) |
| Grounded theory | 20–30 interviews | Theoretical saturation | Charmaz; Glaser & Strauss |
- Effect size: f² = 0.15
- Alpha: α = 0.05
- Power: 1−β = 0.80
- Predictors: k = 5
- Tool: G*Power 3.1
What the justification paragraph should actually say
Three elements, in order: method, number, citation. That's it. Most reviewer-accepted justifications fit into 60–120 words. Length is not what makes it credible — specificity is.
Quantitative template. "A priori power analysis using G*Power 3.1 (Faul et al., 2009) indicated a minimum sample of N = 395 to detect a medium effect (f² = 0.15) at α = .05 with statistical power of 0.80 for a multiple regression with five predictors. The final sample of 412 valid responses exceeded this threshold, yielding observed power of 0.83 for the focal analyses."
Qualitative template. "Sample size was guided by information power (Malterud et al., 2016). Given a narrow study aim, high sample specificity (purposively selected senior clinicians), strong dialogue quality, and a planned thematic analysis, a sample of 18 interviews was judged sufficient. Saturation was monitored iteratively: after interview 16, two additional interviews produced no new themes, confirming adequate information density."
Design-driven template. "Following Yin's (2018) replication logic for multi-case research, three contrasting cases were selected to maximise variation across institutional type. The aim was theoretical generalisation through pattern matching, not statistical generalisation, making three cases sufficient for the cross-case analysis planned."
For survey-based work, "327 responses were collected" is incomplete. Reviewers need the denominator — how many invitations were sent, how many were eligible — to assess non-response bias. The full sentence should be "1,420 invitations were distributed; 412 valid responses were returned (response rate 29%)." Missing the denominator is the single easiest rejection trigger to remove.
Sample size justification is the cheapest insurance policy in your manuscript. A defended paragraph closes an objection that an asserted number opens.
The three mistakes reviewers always catch
1. Asserting adequacy without a method. "The sample was sufficient for the analyses performed" tells reviewers nothing. Replace it with the method, the number, and the citation.
2. Borrowing the wrong approach. Power analysis for a phenomenological study is a category error. Saturation for an SEM is the same error in reverse. The approach must match the design.
3. Treating sample size as a defensive move only. A defended sample size also strengthens your discussion section — it lets you make claims about generalisability or transferability that an asserted sample never could.
When your sample is smaller than the floor
Not every researcher can hit the minimum. Specialist populations, rare conditions, hard-to-reach professionals, and small institutional settings often produce samples below the standard floor. The honest move is to acknowledge it, defend it, and qualify your claims.
State the constraint directly ("recruitment was limited to senior clinicians in two specialist units"), report the achieved sample with full transparency, and explicitly frame the analysis as exploratory rather than confirmatory where appropriate. Use methods designed for smaller samples — PLS-SEM rather than covariance-based SEM, Bayesian estimation with informed priors, bootstrapping for confidence intervals. Reviewers respect honest constraints far more than borderline samples dressed up as adequate ones.
Closing — the rule that solves most of it
For every sample size claim in your methodology, ask three questions: What method led to this number? Is the method appropriate for this design? Have I cited a recognised source for the method? If you can answer all three in two sentences, the section is done. If you can't, the section needs another draft — not more data.
Need help defending your sample size?
Book a free 30-minute consultation. A PhD editor will look at your design and methodology together — and help you write a sample-size paragraph reviewers won't flag.
Book Free Consultation → Or read: Quantitative Research Methodology — What to Report and How