How to Find a Critical Value

Finding a critical value is the step that turns a significance level into a decision. It gives you the cut-off for your hypothesis test: if your test statistic falls beyond it, you reject the null hypothesis; if it stays inside, you fail to reject it. This guide explains how to find a critical value step by step — for z and for t, from a table, in Excel, and on a TI-84 — and finishes with the common values, a worked example, and the mistakes to avoid.

To find a critical value, choose your significance level α (usually 0.05), pick z if the population standard deviation is known or the sample is large, otherwise pick t with df = n − 1, split α in half for two-tailed tests, then read the value from a z- or t-table or compute it in Excel or on a TI-84.

What Is a Critical Value?

A critical value is the cut-off point that separates the rejection region from the rest of the distribution. If your test statistic lands beyond the critical value, the result sits in the rejection region and you reject the null hypothesis. If it lands inside the critical value, the result is not unusual enough to reject the null, and you fail to reject it.

In other words, the critical value is the boundary set by your significance level α. With α = 0.05, you reject the null hypothesis whenever the probability of seeing a result at least that extreme is below 5% if the null were true. For a z-test, that boundary is the critical z-value; for a t-test, it is the critical t-value from the t-distribution.

The same idea underlies confidence intervals: the critical value is the multiplier that stretches the standard error into a margin of error. A 95% confidence interval, for example, is the sample statistic plus or minus the critical value times the standard error.

How to Find a Critical Value in 4 Steps

Finding a critical value is a four-step process. Work through the steps in order, because each one decides the next.

  1. Pick your significance level α. Most tests use α = 0.05. A stricter test uses α = 0.01 and a looser one α = 0.10. Your α is the total size of the rejection region.
  2. Pick the right distribution. Use z when the population standard deviation σ is known, or when the sample is large (n ≥ 30). Use t when σ is unknown and the sample is small. If you are not sure which to use, the comparison in the t-distribution vs normal distribution guide settles it.
  3. Decide one-tailed or two-tailed, and split α if needed. A one-tailed test puts all of α in one tail. A two-tailed test splits it, α/2 in each tail. This is the step most people get wrong — see the mistakes section below.
  4. Look the value up or compute it. Read it from a z- or t-table, get it from Excel, or use a TI-84. The next sections show each method.

How to Read a Z-Table for a Critical Value

A z-table gives cumulative areas under the standard normal curve, so you must convert your α into an area before you look anything up. For a two-tailed test, look up the area 1 − α/2. For a one-tailed test on the right, look up 1 − α; for a one-tailed test on the left, look up α. Then find the z-score that gives that area.

With the usual α = 0.05 for a two-tailed test, look up 1 − 0.05/2 = 0.975. The z-score for a cumulative area of 0.975 is 1.960, so the critical values are ±1.960. For a one-tailed test at α = 0.05, look up 1 − 0.05 = 0.95, which gives 1.645. Our Z-Table Calculator checks the lookup for you, and the Critical Z-Value Calculator gives the value directly.

How to Read a T-Table for a Critical Value

A t-table is organised differently from a z-table: it is indexed by the significance level α and the degrees of freedom, not by cumulative area. That means you do not convert α into an area first — the table does the splitting for you.

To find a t critical value, first work out the degrees of freedom, df = n − 1. Then find the row for that df and the column for your α. For a two-tailed test at α = 0.05, use the column labelled for two tails; for a one-tailed test, use the one-tailed column. The number where the row and column meet is your critical t-value.

With n = 12, df = 11. At α = 0.05, the two-tailed column gives 2.201 and the one-tailed column gives 1.796. If you want the full walkthrough, see how to read a t-table, or compute the value with the T-Distribution Calculator.

The authoritative reference table is the NIST/SEMATECH e-Handbook: Critical Values of the Student’s t Distribution. It lists the critical value for every degrees of freedom and significance level.

How to Find a Critical Value in Excel

Excel’s built-in functions give the critical value directly — no table needed. The formulas use α and, for t, the degrees of freedom.

  • Two-tailed z: =NORM.S.INV(1−α/2) — with α = 0.05, =NORM.S.INV(0.975) returns 1.960.
  • One-tailed z (right): =NORM.S.INV(1−α) — with α = 0.05, =NORM.S.INV(0.95) returns 1.645. For a left tail, take the negative.
  • Two-tailed t: =T.INV.2T(α, df) — with α = 0.05 and df = 11, =T.INV.2T(0.05, 11) returns 2.201.
  • One-tailed t (right): =T.INV(1−α, df) — with α = 0.05 and df = 11, =T.INV(0.95, 11) returns 1.796.

Older functions NORMSINV and TINV still work in most versions of Excel, but NORM.S.INV, T.INV.2T and T.INV are the current ones and the ones to use.

How to Find a Critical Value on a TI-84

A TI-84 (or TI-84 Plus) finds critical values with two inverse functions: invNorm for z and invT for t. They take a cumulative area, so again you must convert α first.

  1. Press 2ND then VARS to open the DISTR menu.
  2. For a z critical value, choose invNorm( and type the area: invNorm(1−α/2) for a two-tailed test, or invNorm(1−α) for a one-tailed test on the right. With α = 0.05, two-tailed: invNorm(0.975) = 1.960.
  3. For a t critical value, choose invT( and type the area and the degrees of freedom: invT(1−α/2, df) for a two-tailed test, or invT(1−α, df) for a one-tailed test on the right. With α = 0.05 and df = 11, two-tailed: invT(0.975, 11) = 2.201.
  4. For a left-tailed test, take the negative of the right-tail value — the distributions are symmetric.

invNorm uses a mean of 0 and a standard deviation of 1 by default, so you can leave μ and σ at their defaults when finding a critical z-value.

Common Critical Values Table

These are the critical values you will use most often. They match the NIST/SEMATECH e-Handbook table, and they are worth keeping close while you work.

Confidence levelz critical value, two-tailedz critical value, one-tailed (right)
90%±1.6451.282
95%±1.9601.645
99%±2.5762.326
Degrees of freedom (df)t critical value, two-tailed, 95%t critical value, one-tailed (right), 95%
92.2621.833
112.2011.796
192.0931.729
292.0451.699

Values are rounded to three decimal places and match the NIST/SEMATECH e-Handbook table. For df = 9 and df = 11, the one-tailed values are the same ones used in the worked example below.

Worked Example: Finding the Critical Value for n = 12

Suppose you test a sample of n = 12 people at a significance level of α = 0.05 for a two-tailed test. The critical value depends on what you know about the population:

  1. The population standard deviation is unknown — use t. With n = 12, the degrees of freedom are df = 12 − 1 = 11. For a two-tailed test at α = 0.05, the table gives t* = 2.201. Reject the null hypothesis if |t| > 2.201.
  2. The population standard deviation is known — use z. The standard normal distribution has no degrees of freedom, so the critical value is the same one you read from a z-table or from Excel: z* = 1.960. Reject the null hypothesis if |z| > 1.960.
  3. A one-tailed (right) version of the t case. With df = 11 and all of α = 0.05 in the right tail, the critical value is t* = 1.796. Reject the null hypothesis if t > 1.796.

Notice how much larger the t critical value is than the z one (2.201 against 1.960). That extra width is the price of estimating σ from a small sample: with only 12 observations the t-distribution is wider than the normal, so you need stronger evidence before rejecting. As the sample grows, the t critical value falls towards 1.960. If you are running the test itself, our one-sample t-test calculator compares your test statistic with t* for you.

Critical Value vs P-Value

A critical value and a p-value answer the same question from opposite sides. The critical value is the cut-off you set before you look at your data: it is fixed by your significance level α and by whether you are using a z- or a t-distribution. The p-value is the evidence your data actually gives you, calculated after your test statistic is in hand.

That is why the two decision rules always agree. You reject the null hypothesis when your test statistic falls beyond the critical value — |t| > 2.201, say — and that happens exactly when the p-value is smaller than α. If your test statistic stays inside the critical value, the p-value is larger than α and you fail to reject.

The mistake to avoid is comparing the two numbers with each other. A p-value is a probability, so it sits between 0 and 1, while a critical value is measured in test-statistic units — 1.960 or 2.201, for example. Comparing a p-value of 0.03 with a critical value of 2.201 means nothing: each belongs to its own comparison. Compare the test statistic with the critical value, and compare the p-value with α.

In practice, set your α first, find your critical value, and the line in the sand is drawn before the data can influence it. Then run your test and read off the p-value — it tells you how far past (or short of) that line your evidence landed.

Common Mistakes When Finding Critical Values

Most errors in finding a critical value come from skipping one of the four steps. Watch for these four:

  1. Forgetting to split α for a two-tailed test. With α = 0.05, the two-tailed critical value is z* = 1.960 (area 0.975), not 1.645. Using the one-tailed value makes the rejection region too wide, so you reject the null hypothesis far more often than 5% of the time.
  2. Using z with a small sample and an unknown σ. With n = 12 and σ unknown you need t, not z: t* = 2.201, not z* = 1.960. The z value is too small, so it rejects too easily. Only use z when σ is known or the sample is large (n ≥ 30).
  3. Mixing up the t-table’s rows and columns. The row is the degrees of freedom, df = n − 1, and the column is the significance level α — not a cumulative area, as on a z-table. Reading the row for the sample size, or the column for the area, gives a different, wrong value.
  4. Confusing the critical value with the p-value. The critical value is the cut-off you fix in advance; the p-value is the evidence your data gives you. Compare your test statistic with the critical value, and your p-value with α — never the p-value with the critical value itself.

What Is the Critical Value for 95% Confidence?

For a 95% confidence level, the critical value depends on whether you are using a z- or a t-distribution, and on whether the test is one- or two-tailed:

  • Two-tailed z: ±1.960 — the value that leaves 2.5% of the standard normal curve in each tail.
  • One-tailed z (right): 1.645 — the value that leaves 5% in the right tail.
  • Two-tailed t: it depends on the degrees of freedom. With df = 11, t = ±2.201; with df = 19, t = ±2.093; with df = 29, t = ±2.045.
  • One-tailed t (right): with df = 11, t = 1.796.

As the sample grows, the t critical value falls towards the z value of 1.960 — for a very large sample the two are effectively the same.

How Do You Find the Critical Value on a TI-84?

On a TI-84 (or TI-84 Plus), you find a critical value with the DISTR menu:

  1. Press 2ND, then VARS, to open the DISTR menu.
  2. For a z critical value, choose invNorm( and type the cumulative area: invNorm(1−α/2) for a two-tailed test, or invNorm(1−α) for a one-tailed test on the right.
  3. For a t critical value, choose invT( and type the area and the degrees of freedom: invT(1−α/2, df) for a two-tailed test, or invT(1−α, df) for a one-tailed test on the right.
  4. With α = 0.05 in a two-tailed test, invNorm(0.975) gives 1.960 and invT(0.975, 11) gives 2.201 (df = 11). For a left tail, take the negative of the right-tail value.

How Do You Find the Critical Value for a Two-Tailed Test?

For a two-tailed test, split α in half and use the area that leaves only α/2 in the tail:

  • z critical value: look up the area 1 − α/2 in the z-table. With α = 0.05, that is 1 − 0.025 = 0.975, which gives z = ±1.960. In Excel, =NORM.S.INV(1−α/2).
  • t critical value: find the row for df = n − 1 and the column for α on a t-table. With α = 0.05 and df = 11, the two-tailed column gives t = ±2.201. In Excel, =T.INV.2T(α, df).

The ± matters: a two-tailed test has a rejection region on both sides, so you reject the null hypothesis if your test statistic is less than −1.960 or greater than +1.960 (for z at α = 0.05).

Frequently Asked Questions

For a two-tailed test at 95% confidence, the z critical value is ±1.960, and for a one-tailed test on the right it is 1.645. For a t critical value, the answer depends on the degrees of freedom: with df = 11, the two-tailed t critical value is ±2.201 and the one-tailed value is 1.796.

Press 2ND, then VARS, to open the DISTR menu. For a z critical value choose invNorm and type the cumulative area: invNorm(1−α/2) for a two-tailed test or invNorm(1−α) for a one-tailed test on the right. For a t critical value choose invT and type the area and the degrees of freedom: invT(1−α/2, df) for two-tailed or invT(1−α, df) for one-tailed. With α = 0.05 and df = 11, invNorm(0.975) = 1.960 and invT(0.975, 11) = 2.201. For a left tail, take the negative of the right-tail value.

Split α in half and use the area 1 − α/2. For a z critical value, look up that area in the z-table: with α = 0.05, 1 − 0.025 = 0.975 gives z = ±1.960. For a t critical value, find the row for df = n − 1 and the α column: with α = 0.05 and df = 11, the two-tailed value is ±2.201. The ± matters: a two-tailed test has a rejection region on both sides, so you reject the null hypothesis if your test statistic is less than the negative critical value or greater than the positive one.

A critical value is a cut-off you fix before the test, from your significance level α and the distribution (z or t). It marks the boundary of the rejection region. A p-value is a probability worked out from your data after the test: the probability of getting a test statistic at least as extreme as the one observed if the null hypothesis were true. The two decision rules always agree — reject the null if the test statistic goes beyond the critical value, which is the same as rejecting when p < α. Never compare the two numbers with each other directly, because one is a z or t value and the other is a probability between 0 and 1.

Use a z critical value when the population standard deviation σ is known, or when the sample is large (n ≥ 30) so the sample standard deviation is a good stand-in. Use a t critical value when σ is unknown and the sample is small (n < 30), with df = n − 1. The t critical value is always larger than the z value for the same confidence level, because it allows for the extra uncertainty in estimating σ — for example, at 95% two-tailed, z* is 1.960 while t* with df = 11 is 2.201. As the sample size grows, t* moves towards z*.

For a two-tailed test at 99% confidence, the z critical value is ±2.576, and for a one-tailed test on the right it is 2.326. For a t critical value, the answer again depends on the degrees of freedom: with df = 11, the two-tailed t critical value is ±3.106 and the one-tailed value is 2.718. The higher confidence level makes the interval wider, so the critical value is larger than the 95% values of ±1.960 (z) and ±2.201 (t, df = 11).

Excel has four functions for this. For a z critical value use =NORM.S.INV(1−α/2) for a two-tailed test or =NORM.S.INV(1−α) for a one-tailed test. For a t critical value use =T.INV.2T(α, df) for a two-tailed test or =T.INV(1−α, df) for a one-tailed test. With α = 0.05 and df = 11, the four functions return 1.960, 1.645, 2.201 and 1.796 in that order. For a left tail, take the negative of the right-tail value.

It changes your decision, and usually for the worse. A critical value that is too small makes you reject the null hypothesis too often, so you get more false positives (Type I errors). One that is too large makes you fail to reject when a real effect is present, so you get more false negatives (Type II errors) and less power. The same error makes a confidence interval too narrow or too wide. Before you decide, check the four things most often got wrong: z or t, one-tailed or two-tailed, α, and the degrees of freedom.

Conclusion

Finding a critical value comes down to four steps: choose your significance level, pick z or t, decide one-tailed or two-tailed, and look the value up or compute it. At 95% confidence, the two-tailed z critical value is ±1.960, and with df = 11 the two-tailed t critical value is ±2.201. Keep the four checks in mind — distribution, tails, α and degrees of freedom — and the cut-off you draw will be the right one.

Check Tools

Try the calculators behind the methods in this guide:

Browse the full set of calculators in Tools.