Three sampling designs
- Simple random — every row has the same probability of selection and rows are sampled without replacement.
- Systematic — pick a random start and then step through the population at a fixed interval. This is efficient for ordered lists, but avoid it when the ordering has a periodic pattern related to the variable you study.
- Stratified — split rows by a column (region, segment, type…) and sample within every group. Proportional allocation preserves population shares; equal allocation deliberately gives small strata more representation.
Reproducible versus unpredictable draws
Leave Seed empty for a fresh draw based on the browser's cryptographic random generator. Enter a seed for an audit, classroom exercise or analysis that another person must reproduce: the same data, method, sample size and seed produce the same selected row numbers.
Sample-size helper
The helper uses the standard normal approximation for a proportion, assumes the worst-case proportion of 50%, and applies the finite-population correction. It is a planning aid, not a guarantee: non-response, clustering, weighting, rare-event estimates and complex survey designs can require a different calculation.
Keep the source row numbers
The result can download both the sampled CSV and the original source row numbers. That makes the selection auditable without modifying the source file. The header is row 1, so the first data row is reported as source row 2.
Is my data uploaded?
No. The sample is drawn in your browser, so nothing you add is sent to gratistools.be. See the privacy page for details.