How to Find Critical Values in Excel [Complete Guide]

“Critical value” is the cutoff point for a hypothesis test. If your test statistic lands past that cutoff (in the right tail or either tail), you reject the null hypothesis.

Excel can compute critical values directly with built-in inverse distribution functions. The key is matching the right distribution and handling tails correctly—most mistakes come from mixing up alpha, confidence level, and one-tailed vs two-tailed probabilities.

As an Amazon Associate I earn from qualifying purchases.

What Critical Values Mean (and why you’ll use Excel)

In classical hypothesis testing, you choose a significance level α (like 0.05). The critical value(s) come from the chosen distribution and α.

Excel’s job is to invert the cumulative distribution function (CDF). Instead of asking “What’s P(X ≤ x)?”, you ask “What x makes P(X ≤ x) equal to a given probability?” That’s exactly what functions like T.INV.2T, NORM.S.INV, CHISQ.INV.RT, and F.INV.RT do.

Before You Start: Inputs Excel Needs

You’ll typically need some or all of the following:

• Distribution type: Normal (Z), t, chi-square, or F.
• Degrees of freedom (df) for t/chi-square/F.
• Significance level α (e.g., 0.01, 0.05, 0.10).
• Tail type: right-tailed, left-tailed, or two-tailed.
• Confidence level if you’re building from CIs (e.g., 95% means α = 0.05).

Excel uses probability areas as inputs. For example, “right-tail critical value” means the value c such that P(X > c) = α.

Quick Map of Critical Values to Excel Functions

Use this table to pick the right function for your distribution and tail.

Distribution	Right-tail (P(X > c) = α)	Left-tail (P(X < c) = α)	Two-tailed (P(|X| > c) = α)
Normal Z (standard)	NORM.S.INV(1-α)	NORM.S.INV(α)	NORM.S.INV(1-α/2)
Normal with mean/std dev	NORM.INV(1-α, μ, σ)	NORM.INV(α, μ, σ)	NORM.INV(1-α/2, μ, σ)
t (df)	T.INV.RT(α, df)	T.INV(α, df)	T.INV.2T(α, df) (returns positive cutoff)
Chi-square (df)	CHISQ.INV.RT(α, df)	CHISQ.INV(α, df)	Typically use two separate one-tailed critical values
F (df1, df2)	F.INV.RT(α, df1, df2)	F.INV(α, df1, df2)	Typically two-tailed uses both tails

Excel’s function naming is consistent: .RT means right-tail, and the base name without .RT corresponds to the left-tail inverse.

Critical Values for the Normal (Z) Distribution

If your statistic follows a normal distribution under H0 and you’re using a Z test (often with known σ), Excel can compute the cutoff quickly.

Standard normal (mean 0, std dev 1)

Use these forms for critical values based on α:

• Right-tailed: =NORM.S.INV(1-α)
• Left-tailed: =NORM.S.INV(α)
• Two-tailed (positive cutoff): =NORM.S.INV(1-α/2)

For two-tailed tests, the negative cutoff is just the negative of the positive cutoff.

Normal with mean μ and standard deviation σ

If your normal variable isn’t standardized, use the non-standard inverse:

• Right-tailed: =NORM.INV(1-α, μ, σ)
• Left-tailed: =NORM.INV(α, μ, σ)
• Two-tailed positive cutoff: =NORM.INV(1-α/2, μ, σ)

Critical Values for the t Distribution

The t distribution is used when σ is unknown and you estimate it (common in one-sample and two-sample t tests). You’ll need df.

Right-tailed t critical value

Use T.INV.RT, where the first argument is the right-tail probability α:

=T.INV.RT(α, df)

This returns the cutoff c such that P(T > c) = α.

Left-tailed t critical value

Use T.INV for the left tail:

=T.INV(α, df)

This returns the cutoff c such that P(T < c) = α.

Two-tailed t critical value

For a two-tailed test, Excel provides T.INV.2T:

=T.INV.2T(α, df)

This returns the positive critical value. The negative cutoff is =-T.INV.2T(α, df).

Critical Values for Chi-Square

Chi-square critical values are common in tests of variance and in goodness-of-fit. Unlike normal/t, the two-tail case is usually handled by computing separate left and right critical values.

Right-tailed chi-square critical value

=CHISQ.INV.RT(α, df)

Equivalent meaning: P(Χ² > c) = α.

Left-tailed chi-square critical value

=CHISQ.INV(α, df)

Equivalent meaning: P(Χ² < c) = α.

Two-tailed chi-square critical values

For a two-tailed interval at significance α, you typically compute:

• Lower critical value: =CHISQ.INV(α/2, df)
• Upper critical value: =CHISQ.INV.RT(α/2, df)

Your acceptance region is between these bounds.

Critical Values for the F Distribution

F critical values show up in variance ratio tests (ANOVA and two-variance tests). As with chi-square, two-tailed problems usually mean two one-tailed cutoffs.

Right-tailed F critical value

=F.INV.RT(α, df1, df2)

This gives the cutoff c such that P(F > c) = α.

Left-tailed F critical value

=F.INV(α, df1, df2)

This gives the cutoff c such that P(F < c) = α.

Two-tailed F critical values (common CI/ratio setups)

When your problem specifies an overall two-sided α, compute:

• Lower critical value: =F.INV(α/2, df1, df2)
• Upper critical value: =F.INV.RT(α/2, df1, df2)

Some textbooks define two-sided rejection regions differently (especially for reciprocal ratios). If your class uses the reciprocal relationship explicitly, follow that convention.

Workflow Templates (copy these patterns)

If you want consistent results across problems, set up cells for α, df, and tail choice, then reference them inside your formula.

Template A: One-tailed test (generic pattern)

Create inputs like α, df (and df1/df2), then use the correct inverse function for the tail.

α  -> A1
df -> A2
(df1, df2) -> A2, A3

Example for t right-tail:

=T.INV.RT($A$1, $A$2)

Template B: Two-tailed t critical value

Use T.INV.2T and reuse the same cell for the negative bound.

=T.INV.2T($A$1, $A$2)
Negative bound: =-B1

Template C: Confidence interval from confidence level

Confidence level CL (like 0.95) implies α = 1 − CL.

=1-$A$1

Then use α in your critical value formulas (e.g., normal two-tailed positive cutoff: =NORM.S.INV(1-α/2)).

Common Use Cases: One-tailed, Two-tailed, Confidence Intervals

Most problems fall into three categories: hypothesis test cutoffs, confidence interval endpoints, and “accept/reject region” boundaries.

One-tailed hypothesis tests

Right-tailed means the rejection region is on large values. Left-tailed means rejection is on small values.

• Z right-tailed: =NORM.S.INV(1-α)
• t right-tailed: =T.INV.RT(α, df)
• Chi-square right-tailed: =CHISQ.INV.RT(α, df)
• F right-tailed: =F.INV.RT(α, df1, df2)

Two-tailed hypothesis tests

“Two-tailed” usually means split α between both tails (α/2 each).

• Z two-tailed positive cutoff: =NORM.S.INV(1-α/2)
• t two-tailed: =T.INV.2T(α, df) (positive cutoff)
• Chi-square and F: compute left and right separately using α/2

Confidence intervals (endpoints are critical values)

For many standard CI formulas, the endpoint uses the same critical value from a corresponding hypothesis test.

Example patterns:

• Mean (large n, known σ) uses Z critical value at α = 1 − CL
• Mean (unknown σ) uses t critical value at α = 1 − CL
• Variance CI uses chi-square critical values (two endpoints)
• Ratio of variances CI uses F critical values (two endpoints)

Troubleshooting and Gotchas

Critical value calculations are straightforward, but small mistakes can flip your cutoff.

Using the wrong tail function

Quick rule: if the problem says “right-tail probability α,” prefer functions ending in .RT for t/chi-square/F.

Example: CHISQ.INV.RT(0.05, df) is not the same as CHISQ.INV(0.05, df).

Confusing α with 1−CL

If your confidence level is 95%, then α = 0.05 (not 0.95). A common safe approach is to compute α directly in a cell.

α cell: =1-CL

Two-tailed z/t sign handling

For normal and t (two-tailed), you usually report both ± critical values. Excel gives you the positive value; apply the negative with =-value.

Degrees of freedom mismatches

Different tests use different df formulas. For example, one-sample t has df = n−1; two-sample t depends on the method (Welch’s df differs from pooled df). If you paste df from a different formula, your critical value will be off.

Function availability and Excel versions

Functions like T.INV.2T, T.INV.RT, CHISQ.INV.RT, and F.INV.RT are standard in modern Excel (Microsoft 365 and Excel 2019+ in typical installations). If you see #NAME?, verify your Excel version and use older equivalents (like TINV/TINV.2T only if your version supports them), or update Excel.

Non-numeric inputs and locale issues

Excel may treat commas vs semicolons differently depending on locale. If your formula fails, try replacing separators or check “Use system separators” settings.

Worked Examples in Excel (ready-to-follow)

These examples show exactly what to type into Excel and what each result means.

Example 1: Two-tailed z critical value at α = 0.05

Find z critical values for a two-tailed test with α = 0.05 (CL = 95%). Positive cutoff is where 95% of the distribution is below it.

1. Put α in cell A1: 0.05
2. Compute positive cutoff in B1: =NORM.S.INV(1-$A$1/2)
3. Compute negative cutoff in C1: =-B1

Your B1 value should be about 1.959964 (depending on rounding).

Example 2: One-tailed t critical value, df = 12, α = 0.01 (right tail)

1. α in A1: 0.01
2. df in A2: 12
3. Right-tail critical value in B1: =T.INV.RT($A$1,$A$2)

That B1 cutoff is the value c such that P(T > c) = 0.01.

Example 3: Two-tailed t critical value, df = 20, CL = 99%

1. CL in A1: 0.99
2. Compute α in A2: =1-$A$1
3. Positive critical value in B1: =T.INV.2T($A$2, 20)
4. Negative critical value in C1: =-B1

This uses α split across two tails automatically via T.INV.2T.

Example 4: Chi-square two-tailed critical values for variance CI

Suppose df = 9 and you want a 95% confidence interval, so α = 0.05 and you use α/2 = 0.025 in each tail.

1. df in A1: 9
2. α in A2: 0.05
3. Lower critical value in B1: =CHISQ.INV($A$2/2, $A$1)
4. Upper critical value in C1: =CHISQ.INV.RT($A$2/2, $A$1)

Your CI acceptance region for Χ² lies between B1 and C1.

Example 5: F right-tail critical value for df1 = 5, df2 = 10, α = 0.05

1. α in A1: 0.05
2. df1 in A2: 5
3. df2 in A3: 10
4. Right-tail cutoff in B1: =F.INV.RT($A$1, $A$2, $A$3)

This is the c such that P(F > c) = 0.05.

Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.Support on Ko-Fi

Comparisons and Alternatives

Excel formulas are usually the fastest and most reproducible option, but it depends on what your assignment expects.

Excel vs Stat tables

Printed critical value tables often round to fewer decimals (e.g., 1.96 instead of 1.959964). Excel gives more precise values, which can slightly change decision boundaries at strict grading thresholds.

Excel vs built-in Data Analysis add-ins

Data Analysis tools (if installed) help with tests but don’t always display critical values directly for every configuration. When you need exact critical cutoffs, inverse distribution functions are more reliable.

Using regression output to get critical t values

Regression output in Excel may report t-statistics and p-values, but the critical values are still based on t distribution with appropriate df. If your workflow requires explicit critical values (for plotting or manual decision), compute with T.INV/T.INV.RT/T.INV.2T.

FAQs

Which Excel function returns the critical value for a two-tailed t test?

Use T.INV.2T(α, df) for the positive critical value. The negative cutoff is the same magnitude with a negative sign.

When I use NORM.S.INV, why do I see 1-α or 1-α/2?

NORM.S.INV(p) expects a left-tail probability p meaning P(Z ≤ x) = p. For a right-tail cutoff you want P(Z > c) = α, which means P(Z ≤ c) = 1 − α. For two-tailed, each tail is α/2.

What does T.INV.RT mean, exactly?

T.INV.RT(α, df) returns the value c such that P(T > c) = α. It directly matches right-tail significance levels.

How do I compute both critical values for a two-tailed chi-square test?

Compute lower with CHISQ.INV(α/2, df) and upper with CHISQ.INV.RT(α/2, df). Then use the interval between them as the acceptance region for Χ².

I’m getting #NUM! or strange results—what should I check?

Confirm df is positive and matches the test. Also ensure α is between 0 and 1 (not a percentage like 5 without converting to 0.05). Finally, verify you used .RT for right-tail cutoffs.

Bottom Line

To find critical values in Excel, pick the right distribution and use the corresponding inverse function, then match the tail definition precisely. For normal and t, you can get two-tailed values directly (normal via NORM.S.INV(1-α/2), t via T.INV.2T), while chi-square and F usually require separate left and right calculations.

Once you standardize your inputs (α, df, tail), your spreadsheet becomes a repeatable reference—not a one-off calculation you have to re-check every time.

Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.