The P-Value Calculator converts a test statistic into a p-value for z, t, chi-square or F tests, with a one-tailed or two-tailed option where the test allows it. Enter the statistic, degrees of freedom when required, and the tail choice; the calculator returns the p-value, compares it with a chosen α, and shades the rejection region on a tail-area plot.
A p-value answers: if the null hypothesis were true, how often would data at least this extreme appear by chance? It does not answer whether the null is true. That distinction drives the misinterpretation section below.
Calculate a p-value from a test statistic
Supported inputs are z, t with degrees of freedom, chi-square with degrees of freedom, and F with numerator and denominator degrees of freedom. Enter the statistic from the test output; the calculator maps it onto the matching reference distribution and returns the tail probability.
For z and t, set one-tailed or two-tailed before reading the number so the shade matches the hypothesis.
For z and t, choose one-tailed or two-tailed before reading the number. Chi-square and F tests are typically one-sided (upper tail) in the forms taught in intro courses; the interface reflects that.
Choose a one-tailed or two-tailed test
A two-tailed test asks whether the statistic differs from the null value in either direction and doubles the smaller one-tail area for symmetric z and t tests. A one-tailed test asks about one named direction only.
Use two-tailed unless the research plan committed to a direction before seeing the data; switching after observing the sign halves p and inflates false positives.
Use two-tailed unless the research plan committed to a direction before seeing the data. Switching to one-tailed after observing the sign halves the p-value and inflates false positives. Directional hypotheses belong in the protocol, not in a post-hoc rescue of a 0.06.
Compare the p-value against alpha
Alpha is the prechosen significance level, often 0.05 and sometimes 0.01 or 0.10. If p is less than or equal to alpha, reject the null hypothesis; if p is greater than alpha, fail to reject it. Fail to reject is not the same as proving the null true.
The study may simply lack power, which is why intervals and effect sizes belong beside any binary decision.
- If p ≤ α, reject the null hypothesis.
- If p > α, fail to reject the null hypothesis.
"Fail to reject" is not the same as "prove the null." The study may simply lack power. The calculator highlights the comparison once you set α beside the result.
Calculate a two-tailed p-value from z equals 2.0
For z equals 2.0 in a two-tailed normal test, the one-tail area above 2.0 is about 0.0228. Doubling that area produces a two-tailed p-value of about 0.0455. Against alpha 0.05 that result rejects the null; against alpha 0.01 it does not. The data did not change between those two decisions; only the threshold did.
1. One-tail area above 2.0 ≈ 0.0228 (more precisely 0.02275). 2. Two-tailed p = 2 × 0.02275 ≈ 0.0455. 3. Against α = 0.05: 0.0455 < 0.05 → reject the null.
Against α = 0.01 the same p-value fails to reject. The data did not change; the threshold did. That is why α must be set before the look at p, not negotiated afterward.
Read the tail area plot
The plot shows the reference density, a vertical line at the test statistic, and shaded rejection regions for the chosen α. For two-tailed α = 0.05 on a normal curve, critical values sit near ±1.96. A statistic beyond either critical value falls in the shade.
Check that the shaded tails match the selected test direction. A large negative statistic with only an upper-tail shade selected is a setup error, not evidence against the null. The critical-value markers at plus or minus 1.96 for two-tailed alpha 0.05 give a second visual check against the numeric p.
If the statistic line sits inside the unshaded centre while p is still reported, the visual and the number agree: fail to reject at that α. Use the plot to sanity-check sign errors (a large negative z with only an upper-tail shade selected).
Understand what a p-value does not mean
Five misreadings dominate published mistakes, and each is wrong for a specific reason spelled out below. A p-value is a tail probability under the null model, not a direct probability that a hypothesis is true, not an effect size, and not a replication chance. The 0.05 cutoff is a convention, and a non-significant result is not proof of no effect.
1. A p-value is not the probability that the null hypothesis is true. It is the probability of data at least this extreme *if* the null were true. Those are different conditionings. Confusing them turns a tail probability into a direct bet on the hypothesis.
2. A p-value is not an effect size. A trivial difference reaches p < 0.05 with a large enough sample. Separately report a mean difference, risk ratio or standardised effect; do not treat small p as "large effect."
3. A p-value is not the probability of replication. A p of 0.03 does not mean a repeat study has a 97% chance of finding the same result. Replication depends on power, true effect size and study design.
4. The 0.05 threshold is a convention, not a law of nature. p = 0.049 and p = 0.051 are nearly identical evidence. Dichotomising them into "significant" and "not" for journal habit does not create a bright line in the data.
5. Non-significant is not "no effect." Failing to reach α may mean the study was underpowered. Absence of evidence is not evidence of absence; check confidence intervals and power before claiming a null finding.
These points align with the American Statistical Association's 2016 statement on p-values. The calculator still compares p with α because that workflow is what coursework and many protocols require; the section above is what keeps the number from being over-read.
Read the critical value for a chosen alpha
Critical values are the cutoffs on the statistic scale that match a chosen alpha and tail rule. For a two-tailed normal test they are about 1.645 at 0.10, 1.960 at 0.05 and 2.576 at 0.01. Reject when the absolute z exceeds the critical value, which is equivalent to p less than or equal to alpha.
For t tests the cutoff also depends on degrees of freedom.
| α | Critical | z | |
|---|---|---|---|
| 0.10 | 1.645 | ||
| 0.05 | 1.960 | ||
| 0.01 | 2.576 |
Reject when |z| exceeds the critical value (equivalently when p ≤ α). For t tests the critical value depends on df and is larger than z for small samples. The Confidence Interval Calculator uses the same critical values when building intervals.
Reporting p-values without the usual traps
Report the exact p when software provides it, not only a star or a "p less than 0.05" badge. Pair it with an effect size or a confidence interval so readers can judge magnitude. Precommit to alpha and to one-versus two-tailed rules in the analysis plan when the work is confirmatory.
Exploratory analyses still produce p-values, but they should be labelled as exploratory. Multiple unadjusted tests inflate the chance of at least one false positive; that problem is not solved by this calculator, and it is not solved by quietly switching from two-tailed to one-tailed after seeing the sign.
Frequently asked questions
What is a p-value?
A p-value is the probability, under the null hypothesis, of obtaining a test statistic at least as extreme as the one observed. Smaller p means the data are less compatible with the null under that model.
What is alpha?
Alpha (α) is the prechosen significance level used as a reject/fail-to-reject cutoff, commonly 0.05. It is a policy choice for error control, not a discovered constant.
Should I use a one-tailed or two-tailed test?
Use two-tailed unless a directional hypothesis was specified before seeing the data. One-tailed tests are easier to "pass" and are often misused after the fact.
Is p = 0.049 meaningfully different from p = 0.051?
No. They represent nearly the same strength of evidence. The bright line at 0.05 is conventional.
Does a significant p-value prove the research hypothesis?
No. It indicates data unusual under the null at the chosen α, given the model assumptions. Confounding, bias and wrong models remain possible.
Does a non-significant result prove no effect?
No. The study may lack power. Report intervals and consider effect sizes before claiming a null conclusion.
Which test statistic should I enter?
Enter the statistic your procedure produced: z for large-sample normal tests, t for mean tests with estimated σ, chi-square for many categorical tests, F for many variance-ratio tests. Match df to the same output.
How is a two-tailed p-value computed from z?
Find the one-tail area beyond |z|, then double it. For z = 2.0, one-tail ≈ 0.0228 and two-tail ≈ 0.0455.
Can software disagree slightly on p?
Yes. Tail algorithms and rounding differ at the fourth decimal. Disagreements in the first two decimals usually mean wrong tails, wrong df or the wrong distribution family.
Summary
The P-Value Calculator turns z, t, chi-square or F statistics into p-values, compares them with α and shades the rejection region. A two-tailed z = 2.0 yields p ≈ 0.0455, which rejects at 0.05 and not at 0.01.
A p-value is not the probability the null is true, not an effect size, not a replication probability; 0.05 is a convention; and non-significance is not proof of no effect.
Use this page for the significance decision; use the Z-Score Calculator when you only need the standardised score and curve area.