How this instrument works
The sample proportion, written p-hat (p̂), is the simplest possible statistic: the number of 'successes' in your sample (x) divided by the total number of observations (n). A success just means whatever outcome you're counting — 'voted yes,' 'passed inspection,' 'clicked the link' — not necessarily a good outcome. If 45 out of 150 survey respondents said yes, p-hat is 45/150 = 0.3, or 30%.
p-hat is an estimate, not the truth — it's your sample's best guess at an unknown population proportion (p), the true fraction across everyone you didn't sample. Draw a different sample of the same size from the same population and you'll typically get a slightly different p-hat, just by chance. That sample-to-sample wobble is exactly what this site's sampling error instrument measures, and it's why a single poll result is always reported with a margin of error rather than as a bare percentage.
p-hat is the input every proportion-based inferential tool builds on: standard error of a proportion, confidence intervals for a proportion, and proportion-based hypothesis tests all start by computing x/n first. It looks similar to a basic probability calculation, but the distinction matters — a theoretical probability is derived from known odds, like a fair coin's 50%, while p-hat is an empirical estimate pulled from actual observed data, and it carries sampling uncertainty that a theoretical probability doesn't.
- Enter how many outcomes counted as a success into Number of successes (x) — for example, how many respondents answered yes.
- Enter your total sample count into Sample size (n) — everyone you sampled, successes and non-successes together.
- Read Sample proportion (p-hat) — the fraction of your sample that counted as a success, expressed as a decimal.
- x cannot exceed n — successes are a subset of your total sample, so the instrument will flag an entry where x is larger than n.
Worked example — 45 successes out of 150
A researcher surveys 150 people and finds that 45 of them answered yes to a question. Enter 45 into Number of successes (x) and 150 into Sample size (n).
Sample proportion (p-hat) reads 45 / 150 = 0.3, or 30%. That 0.3 is the point estimate for the true population proportion — the single best guess based on this sample — and it becomes the input for follow-up calculations like the standard error of the proportion, which this site's sampling error instrument computes from this same 0.3 and 150, or a confidence interval built around it.
Questions
Is p-hat the same thing as a percentage?
Essentially yes — p-hat is just expressed as a decimal fraction rather than a percentage. A p-hat of 0.3 and '30%' describe the identical result; multiply p-hat by 100 to get the percentage form most people are used to seeing in poll headlines and news reports.
What's the difference between p-hat and the true population proportion p?
p-hat is what you actually calculated from your sample; p, without the hat, is the unknown true value across the entire population, which you almost never get to observe directly. p-hat is your best estimate of p, and it will typically differ slightly from the true p just because of which specific individuals ended up in your sample — that gap is sampling error.
Can p-hat ever be greater than 1 or less than 0?
No. Because x, the count of successes, can be at minimum 0 and at maximum n (every observation counting as a success), p-hat = x/n is always confined to the range 0 to 1. A calculated value outside that range signals a data-entry mistake, most often entering more successes than the total sample size.
How precise is p-hat as an estimate of the true proportion?
That depends on your sample size — precision isn't something p-hat itself reports. A p-hat of 0.3 from a sample of 150 is more precise than a p-hat of 0.3 from a sample of 15, even though both give the identical point estimate. This site's sampling error instrument quantifies that precision directly, using both p-hat and n as inputs.
How is p-hat different from a straightforward probability calculation?
A theoretical probability, like a fair die landing on 4 (1/6), is derived from known rules or symmetry before any data is collected. p-hat is an empirical estimate pulled from actually observed outcomes, and it carries sampling uncertainty — a different sample would likely produce a slightly different p-hat, whereas a theoretical probability doesn't change from one calculation to the next.