Skip to content
K Knidox Search…
Statistics · Hypothesis testing

Critical Value Calculator

Exact critical values for the z, t, χ² and F distributions from your α and degrees of freedom.

Distribution
Significance level
Test type
n − 1 for a one-sample test.
Critical value
t = ±2.0860

Reject H₀ when |t| > 2.0860 at df = 20.

A critical value is the cut-off that separates the rejection region from the rest of a distribution. For a two-tailed z-test at α = 0.05 it is ±1.96: any test statistic beyond that rejects the null hypothesis, because only 5% of the curve lies outside those bounds.

What a critical value is

Every hypothesis test produces a statistic and needs a threshold to judge it against. The critical value is that threshold. You choose a significance level α — the risk of a false positive you are willing to accept — and the critical value is the point with exactly α of the distribution beyond it. Land past it and the result is significant; land short and it is not.

This is the mirror image of the p-value approach. A p-value asks how much of the distribution lies beyond the statistic you actually got; a critical value fixes the area first and asks where the boundary sits. The two always agree, because p < α happens precisely when the statistic passes the critical value. Which one you report is convention, not mathematics.

Choosing the distribution

Use z when the population standard deviation is known or the sample is large. Use t when σ is estimated from the sample, which needs degrees of freedom. Use χ² for tests on counts and on a single variance, and F for comparing variances or running an ANOVA — F needs two separate degrees-of-freedom values, one for each variance in the ratio.

P(X > critical value) = α

for a two-tailed test the area α is split evenly, leaving α ÷ 2 in each tail

  1. 1
    Pick the distribution the test uses. A two-sample comparison of means with an estimated standard deviation uses t, not z.
  2. 2
    Set the significance level. α = 0.05 is the usual default and corresponds to a 95% confidence level.
  3. 3
    Decide how many tails the test has. A directional hypothesis is one-tailed and keeps all of α at one end; a non-directional one splits it, leaving 0.025 in each tail at α = 0.05.
  4. 4
    Supply the degrees of freedom. Comparing two groups of 11 with pooled variance gives df = 11 + 11 − 2 = 20.
  5. 5
    Read off the cut-off and apply it. At df = 20, two-tailed α = 0.05, the critical value is ±2.086, so a calculated t of −2.31 falls in the rejection region.

Critical values you will meet constantly

z values are exact; t is shown at df = 20, χ² at df = 1, and F at df₁ = 2, df₂ = 27.

Significance levelz (two-tailed)t, df = 20 (two-tailed)χ², df = 1F, df = (2, 27)
α = 0.10±1.645±1.7252.7062.511
α = 0.05±1.960±2.0863.8413.354
α = 0.01±2.576±2.8456.6355.488

Reading the result correctly

Passing the critical value means the data would be unusual if the null hypothesis were true — nothing more. It is not the probability that the null is false, and it says nothing about how big the effect is. A tiny, meaningless difference clears any critical value once the sample is large enough, which is why an effect size belongs next to every significant result.

Two practical cautions. Choose the number of tails before seeing the data: switching to a one-tailed test afterwards because the two-tailed result missed doubles your real false-positive rate. And when you run several tests on the same data, the chance that at least one clears its critical value by luck grows quickly, so adjust for it with a Bonferroni correction or an equivalent procedure.

What is the critical value for a 95% confidence level?
For a z-based interval it is ±1.96, because 95% of the standard normal curve lies between those bounds. With a t-distribution the value depends on degrees of freedom — ±2.086 at df = 20, approaching 1.96 as the sample grows.
What is the difference between a critical value and a p-value?
A critical value fixes the area α first and reports the boundary; a p-value starts from your test statistic and reports the area beyond it. They always reach the same conclusion, since p < α is exactly the condition for passing the critical value.
How does a one-tailed critical value differ from a two-tailed one?
A one-tailed test puts all of α in a single tail, giving a smaller cut-off — 1.645 for z at α = 0.05. A two-tailed test splits α between both tails, pushing the cut-off out to 1.96 for the same significance level.
Why do χ² and F have no ± version?
Both statistics are built from squared quantities, so neither can be negative and both tests reject only on large values. That makes them inherently one-tailed, with a single upper cut-off rather than a symmetric pair.
Which degrees of freedom do I enter?
For a one-sample t-test use n − 1; for a pooled two-sample test use n₁ + n₂ − 2. For χ² use categories − 1, or (rows − 1) × (columns − 1) for a contingency table. For F enter the numerator df first, then the denominator df.
Does a larger critical value mean a stronger result?
No — it means a stricter threshold. Lowering α from 0.05 to 0.01 raises the cut-off and makes significance harder to reach. The strength of a result comes from the effect size and the confidence interval, not from the size of the cut-off.