P-value Calculator

A p-value answers one specific question: if there were truly no effect or no difference at all, how likely would you be to see a result at least as extreme as the one you actually observed? A small p-value means your result would be unusual under that “no effect” assumption — evidence against it. A large p-value means your result is unremarkable even if nothing real is going on. This calculator converts a test statistic (z) directly into that probability, for one-tailed or two-tailed tests, and tells you whether the result clears your chosen significance threshold.

How the P-value Calculator Works

Enter a z-statistic, choose whether your test is two-tailed, left-tailed, or right-tailed, and set your significance level (commonly 0.05). The calculator computes the corresponding p-value using the standard normal distribution, then compares it directly against your significance level to state a decision: reject the null hypothesis, or fail to reject it. This tool takes the z-statistic as a direct input — it does not calculate one from raw sample data itself, so you’ll typically need to compute that z-value first from your actual data before bringing it here.

Getting a Z-Statistic Before You Start

If you have raw data — a sample mean you want to compare against a known or hypothesized value — you need to convert that into a z-statistic first. The Z-Score Calculator performs exactly that conversion, taking a raw value, a mean, and a standard deviation and returning the corresponding z-score, which then plugs directly into this calculator’s first field. Think of the two tools as sequential steps in the same workflow: Z-Score Calculator converts your data into a standardized statistic, and this calculator converts that statistic into a formal probability and decision.

One-Tailed vs. Two-Tailed: The Setting That Changes Your p-value the Most

This single choice affects the result more than almost any other input. A two-tailed test asks whether your result differs from expectation in either direction — higher or lower — and is the appropriate default whenever you don’t have a strong prior reason to expect the effect in one specific direction only. A one-tailed test (left or right) asks whether your result differs in one specific, pre-specified direction, and produces exactly half the p-value of the equivalent two-tailed test for the same z-statistic — meaning it’s easier to reach statistical significance with a one-tailed test, which is precisely why choosing it after already seeing your data (rather than before, as part of your original design) is considered a serious methodological problem. The direction of your test must be decided based on your hypothesis, not chosen after the fact to produce a more favorable result.

Significance Level: The Line You Draw Before You Look

The significance level, alpha (α), is the threshold you decide on before running your test — commonly 0.05, sometimes 0.01 for situations demanding stronger evidence. If your calculated p-value falls below alpha, the result is considered statistically significant at that level, meaning it would be considered too unusual to attribute to chance alone under the null hypothesis. This calculator applies your chosen alpha automatically and states the resulting decision directly, rather than leaving you to compare the two numbers yourself.

What a Small P-value Does Not Mean

A small p-value is one of the most consistently misinterpreted numbers in applied statistics, so it’s worth being precise. A p-value below 0.05 does not mean there’s a 95% probability your finding is real, and it does not mean there’s only a 5% chance the null hypothesis is true. It means: assuming the null hypothesis were true, there would only be a 5% chance of observing a result this extreme or more extreme purely by chance. That’s a statement about the data given the null hypothesis, not a statement about the probability of the null hypothesis given the data — a subtle but important distinction that trips up even experienced researchers.

P-values and Confidence Intervals Are Two Views of the Same Test

A p-value and a confidence interval are mathematically connected: a 95% confidence interval that excludes your hypothesized value corresponds exactly to a p-value below 0.05 in a two-tailed test at the same significance level. If you’d rather see your result expressed as a range of plausible values instead of a single probability and pass/fail decision, the Confidence Interval Calculator presents the same underlying statistical evidence in that alternative, often more interpretable format.

This Calculator Uses the Normal (Z) Distribution

Like several other statistical tools on this site, this calculator’s p-values are based on the normal distribution, which is standard and reliable for large samples but becomes a slight approximation for small ones. Formal hypothesis testing on small samples (commonly under 30 observations) with an estimated standard deviation typically uses the t-distribution instead, which has heavier tails to account for the added uncertainty of estimating variability from limited data. If your sample size is small, treat the p-value here as a close approximation, understanding that a stricter t-test would produce a slightly larger, more conservative p-value.

Where Your Data Needs to Come From First

Before you can compute a z-statistic at all, you typically need basic descriptive numbers from your raw sample — a mean and a standard deviation at minimum. The Standard Deviation Calculator and the more complete Statistics Calculator both compute those foundational figures directly from a list of numbers, which then feed into the Z-Score Calculator to produce the z-statistic this tool ultimately needs.

Deciding How Much Data You Need Before You Test Anything

Hypothesis testing only works reliably with an adequately sized sample — too few observations, and even a real effect may fail to reach statistical significance simply due to insufficient data. If you’re designing a study or survey and want to know how large a sample you need before collecting any data, the Sample Size Calculator answers that planning question directly, using your target confidence level and margin of error.

The Probability Rules Underneath Hypothesis Testing

P-values are ultimately a specific application of probability theory — the chance of an extreme outcome under a stated assumption. If you want to understand the broader probability rules this kind of calculation is built on — combining, comparing, and interpreting probabilities of independent events — the Probability Calculator covers those foundational concepts more generally.

Frequently Asked Questions

What does a p-value of 0.05 actually mean?

It means that, assuming the null hypothesis is true, there is only a 5% chance of observing a result this extreme or more extreme purely by random variation. It does not mean there’s a 5% chance the null hypothesis itself is true, or a 95% chance your finding is real.

What’s the difference between a one-tailed and two-tailed test?

A two-tailed test checks for a difference in either direction and is the standard default. A one-tailed test checks for a difference in one specific, pre-determined direction only, and produces a p-value exactly half that of an equivalent two-tailed test — but the direction must be chosen before seeing the data, not after.

Do I need raw data to use this calculator?

Not directly — this calculator takes an already-computed z-statistic as its input. If you’re starting from raw sample data, use the Z-Score Calculator first to convert your data into a z-statistic, then bring that value here.

What’s the difference between significance level and p-value?

Significance level (alpha) is a threshold you set in advance, before running your test — commonly 0.05. The p-value is the result your test actually produces. You compare the p-value against alpha afterward to decide whether the result is statistically significant.

Is a p-value calculated with the z-distribution accurate for small samples?

It’s a close approximation, but formal small-sample testing (typically under 30 observations) usually relies on the t-distribution instead, which produces slightly wider, more conservative results to account for the added uncertainty in estimating variability from limited data.