Nursing sample size and power analysis support helps students justify a participant target from the research question, study design, primary outcome and planned analysis rather than using a convenient rule such as “100 participants is enough.” The calculation should be reproducible, based on defensible assumptions and reported at the level the design supports.
SPSS and G*Power can support many common power calculations, but software does not decide which model is appropriate and no single package covers every nursing study design. Some projects need power-based planning; others need precision-based calculations, design-effect adjustments or more specialised methods.
What sample-size and power support can cover
- Identifying the primary analysis that should drive the calculation.
- A-priori sample-size and power planning.
- Effect-size assumptions and their evidence base.
- G*Power or appropriate IBM SPSS power procedures.
- Precision-based sample-size calculations for means or proportions.
- Attrition, unequal allocation and clustering adjustments where relevant.
- Sensitivity analysis when recruitment capacity is fixed.
- Methods-ready reporting of assumptions and results.
Start with the research question and primary analysis
The participant target depends on what the study is designed to estimate or test. A comparison of two independent groups, paired pre-post study, correlation, chi-square analysis, regression model and prevalence survey require different inputs.
| Research purpose | Possible analysis | Planning inputs |
|---|---|---|
| Compare two independent groups | Independent-samples t-test or another suitable comparison | Expected difference/effect size, variability, alpha, power and allocation ratio |
| Assess change in the same participants | Paired or repeated-measures analysis | Expected change, variability and within-person relationship |
| Assess association between categorical variables | Chi-square or related model | Expected proportions/effect size, table structure, alpha and power |
| Model a continuous outcome | Regression | Expected model effect, number of predictors, alpha and power |
| Estimate prevalence or a mean precisely | Precision-based design | Expected prevalence/variability, confidence level and acceptable margin of error |
The calculation should be linked to the study’s primary objective. Running separate power calculations for every secondary outcome and choosing the most convenient result is not a defensible planning strategy.
Choose power-based or precision-based planning appropriately
Power analysis is appropriate when the study is designed to test a defined hypothesis or detect an effect of specified magnitude. Precision-based planning is often more appropriate when the primary aim is to estimate a prevalence, proportion, mean or other parameter to a desired confidence-interval width.
These are different design goals. A survey estimating prevalence should not automatically be forced into a hypothesis-testing power calculation simply because the software has a power menu.
Justify the effect-size assumption
The effect size may come from closely related prior evidence, a pilot, a clinically meaningful difference or a transparent range of plausible values. Do not silently insert a conventional “small,” “medium” or “large” effect solely to obtain an achievable sample.
Record where the assumption came from and whether the source population, intervention, outcome and design are sufficiently similar to the proposed study. Where uncertainty is substantial, show how the required sample changes across plausible assumptions.
For deeper explanation, see our effect-size guide for nursing research.
Use G*Power or SPSS only when the procedure matches the design
G*Power supports several common test families and remains widely used for a-priori power and sample-size calculations. Current IBM SPSS Statistics also includes dedicated power-analysis procedures for a range of tests, including means, correlations, proportions and regression models.
The correct sequence is question → design → primary analysis → assumptions → software procedure, not software first. If the required design is not represented adequately by a standard procedure, another statistical method may be needed.
For software-specific steps, use our G*Power guide for nursing research. For broader statistical planning, see statistical-test selection for nursing dissertations.
Distinguish a-priori planning, sensitivity analysis and post-hoc power
| Analysis | Useful purpose | Important caution |
|---|---|---|
| A-priori power analysis | Estimate the sample required before data collection | Depends on defensible effect and design assumptions |
| Sensitivity analysis | Show what effect size a fixed feasible sample could detect at specified alpha/power | Does not remove the limitation of a constrained sample |
| Post-hoc observed power | Sometimes generated after a completed study | Should not be used to interpret whether a non-significant result means “no effect” |
After a study is complete, effect estimates, confidence intervals and the design limitations are generally more informative than observed post-hoc power for interpreting results. A retrospective calculation should not be rewritten as though it had been planned before recruitment.
Account for attrition correctly
Distinguish the number of complete analysable cases required from the number that must initially be recruited. If 120 complete cases are required and 15% attrition is expected, the initial target is approximately 120 ÷ 0.85 = 142, not 120 plus 15%.
Attrition assumptions should reflect the population, follow-up period and study context rather than an arbitrary percentage.
Consider unequal groups and clustering
If participants are allocated unequally, fewer observations in one group can reduce efficiency and increase the total sample required. Clustered designs—such as patients nested within wards, practices or schools—also reduce the amount of independent information compared with simple random sampling.
Where clustering matters, the planning may need an intracluster-correlation assumption and design-effect adjustment or a model-specific approach. A simple two-group calculation should not be presented as adequate for a cluster design.
Regression planning needs more than a fixed “cases per predictor” rule
Rules such as 10, 15 or 20 participants per predictor can be convenient heuristics but do not replace a design-specific justification. Required sample size depends on the model, anticipated effect, number and form of predictors, event frequency for categorical outcomes, collinearity and intended precision.
State which model is primary and which predictors are genuinely planned rather than inflating the model with every available variable.
Feasibility should be reported honestly
A statistically desirable sample may be impossible within the dissertation timetable or available population. If the feasible sample is smaller than the planned target, do not reduce the target retrospectively by changing assumptions merely to make the project appear adequately powered.
A sensitivity analysis can show what magnitude of effect the feasible sample could detect and the limitation can be discussed transparently. Alternatively, the study may need a narrower question, simpler model or different design approved through the appropriate academic process.
Sample size does not repair sampling bias
A large convenience sample can still be systematically unrepresentative. Power concerns the ability of a statistical test to detect an assumed effect under specified conditions; it does not guarantee valid recruitment, measurement or causal inference.
Sampling strategy, eligibility, non-response, missing data and measurement quality should therefore be justified alongside the numerical target.
Report the calculation so another reader can reproduce it
A methods paragraph should normally state the primary analysis, effect-size or precision assumption, source of that assumption where relevant, alpha, target power or confidence level, allocation ratio, software/procedure and any adjustment for attrition or design structure.
A weak statement such as “G*Power indicated 100 participants” hides the assumptions that produced the number. A stronger account explains them.
Related nursing statistics resources
- How to calculate sample size for a nursing dissertation.
- G*Power for nursing research.
- Statistical power in nursing research.
- Effect size in nursing research.
- Nursing dissertation data analysis and SPSS support for analysis after the design is established.
These pages intentionally solve different problems rather than repeating one “sample size” article under several keyword variants.
What to send
- The research question, aim and primary objective.
- Study design and planned comparison/model.
- Primary outcome and measurement scale.
- Relevant prior studies, pilot data or clinically meaningful effect assumptions.
- Expected allocation ratio, attrition or clustering information where relevant.
- University or supervisor requirements.
- Maximum feasible recruitment if already constrained.
Statistical and academic integrity
Support can calculate or review a sample-size justification, explain assumptions and run sensitivity checks. It should not choose implausible assumptions merely to create a convenient target, disguise an underpowered study or retrospectively present a calculation as pre-specified when it was not.
Students remain responsible for the final study design, ethics/governance process and recruitment feasibility.
Frequently asked questions
Can SPSS calculate sample size?
Current IBM SPSS Statistics includes power-analysis procedures for several common statistical tests. The procedure must still match the proposed design and analysis.
Is 100 participants enough?
There is no universal threshold. The answer depends on the study objective, analysis, effect or precision, design structure and missing-data expectations.
What if I can recruit only a fixed number?
A sensitivity analysis can show the effect size detectable under the feasible sample and specified power/alpha. The recruitment constraint should still be reported as a limitation where relevant.
Should I calculate post-hoc power after a non-significant result?
Observed post-hoc power is generally not a useful way to interpret a completed study. Focus instead on effect estimates, confidence intervals, study design and uncertainty.
Request nursing sample-size and power support
Send the research question, design, planned analysis and assumptions through the order page. If the primary statistical test is not yet clear, begin with the quantitative nursing dissertation support or main SPSS/data-analysis hub.
References
- Faul, F., Erdfelder, E., Lang, A.-G., & Buchner, A. (2007). G*Power 3: A flexible statistical power analysis program for the social, behavioral, and biomedical sciences. Behavior Research Methods, 39, 175–191.
- Faul, F., Erdfelder, E., Buchner, A., & Lang, A.-G. (2009). Statistical power analyses using G*Power 3.1. Behavior Research Methods, 41, 1149–1160.
- Kang, H. (2021). Sample size determination and power analysis using the G*Power software. Journal of Educational Evaluation for Health Professions, 18, 17.
- Lakens, D. (2022). Sample size justification. Collabra: Psychology, 8(1), 33267.