TINV in Apple Numbers: Two-Tailed Student t Critical Values
Calculate Student t critical values in Apple Numbers using TINV, with sample confidence intervals, two-sided probabilities and troubleshooting.
Verification status: Official documentation referenced; example outcomes independently reasoned or arithmetically checked where stated. Not executed in Apple Numbers. How formula examples are checked
TINV supplies a positive t-distribution cutoff for a probability and degrees of freedom. It is useful when constructing a two-sided Student t test decision boundary or a classical confidence interval that uses an estimated standard deviation. Apple documents exactly two required numeric arguments in its Formulas and Functions Help 15.4. [A1]
Evidence boundary: Apple’s own example appears below. Additional numeric results and interval calculations use independent Python/SciPy mathematics; they are expected values, not demonstrations of actual Apple Numbers execution.
Syntax: order and restrictions
=TINV(probability, degrees-freedom)
| Required argument | Type | Apple-documented rule |
|---|---|---|
probability |
Number | Strictly greater than 0 and less than 1. For the positive t cutoff, the numeric example and inverse check establish that this is a combined two-tailed probability, not an ordinary lower-tail percentile. |
degrees-freedom |
Number | Must be at least 1. For a simple one-sample mean interval with n observations, df is n−1. Different statistical procedures require their own df. |
Both arguments are mandatory: no optional argument and no default is documented. The probability bound excludes exactly 0 and 1, and Apple does not specify any integer rounding/truncation rule for fractional degrees of freedom. [A1, M1]
The English guide uses commas as argument separators. Apple’s overview says a document configured with a comma decimal separator instead uses semicolons, such as =TINV(0,05;10) in an applicable locale. [A3]
Why TINV is two-tailed
Apple’s example =TINV(0.88,2) returns 0.170940864689457. The independent Student t quantile relation that reproduces this is:
P(|T₂| ≥ t) = 0.88, so t = t₂ inverse survival(0.88 / 2).
Thus, for 0 < probability < 1, TINV gives a nonnegative two-sided critical t value. It is not the inverse of the standard one-sided cumulative distribution at p=0.88. This interpretation is grounded in the Apple example and mathematical cross-check; Apple uses the briefer phrase “probability associated with the distribution.” [A1, M1]
Reproduce results in a Numbers table
Create a table named TCritical, with one header row. Fill A2:B5 with the numbers below. Enter =TINV(A2,B2) in C2, then copy down through C5. Use at least six decimals for the display if comparing to the table. [A1, A4]
| Row | A: Two-tail probability | B: Degrees of freedom | C: Formula / independently expected critical t |
|---|---|---|---|
| 2 | 0.88 |
2 |
=TINV(A2,B2) → 0.170940864689457 (Apple-published) |
| 3 | 0.05 |
10 |
=TINV(A3,B3) → 2.22813885198628 |
| 4 | 0.10 |
9 |
=TINV(A4,B4) → 1.83311293265624 |
| 5 | 0.05 |
4 |
=TINV(A5,B5) → 2.77644510519779 |
A two-sided significance level of 5% is supplied as 0.05 (not 95 and not 0.95). For confidence level C, the corresponding two-sided tail probability is 1−C; for a 95% interval use =TINV(1-0.95, df). [M1]
Advanced: a reproducible 95% confidence interval for a mean
Suppose your Measurements table contains five observations in B2:B6:
| Cell | Observed value |
|---|---|
| B2 | 12 |
| B3 | 15 |
| B4 | 11 |
| B5 | 14 |
| B6 | 13 |
In a separate summary table named Results, add these formulas referencing Measurements by Numbers’ documented TableName::Range notation:
| Output | Numbers formula | Independently calculated result |
|---|---|---|
| n | =COUNT(Measurements::B2:B6) |
5 |
| Mean | =AVERAGE(Measurements::B2:B6) |
13 |
| Sample standard deviation | =STDEV(Measurements::B2:B6) |
1.58113883008419 |
| Degrees of freedom | =COUNT(Measurements::B2:B6)-1 |
4 |
| Positive two-tail critical t | =TINV(0.05,COUNT(Measurements::B2:B6)-1) |
2.77644510519779 |
| Margin of error | =TINV(0.05,COUNT(Measurements::B2:B6)-1)*STDEV(Measurements::B2:B6)/SQRT(COUNT(Measurements::B2:B6)) |
1.96324316147756 |
| Lower endpoint | =AVERAGE(Measurements::B2:B6)-TINV(0.05,COUNT(Measurements::B2:B6)-1)*STDEV(Measurements::B2:B6)/SQRT(COUNT(Measurements::B2:B6)) |
11.0367568385224 |
| Upper endpoint | =AVERAGE(Measurements::B2:B6)+TINV(0.05,COUNT(Measurements::B2:B6)-1)*STDEV(Measurements::B2:B6)/SQRT(COUNT(Measurements::B2:B6)) |
14.9632431614776 |
The conventional 95% t interval is mean ± t critical × (sample SD / √n); using the sample SD rather than population SD is intentional. It describes uncertainty under appropriate sampling/model assumptions, not a guarantee that 95% of the individual observations lie in that interval. The numeric results were independently validated, not run in Numbers. [M2]
Names such as Measurements and Results are examples; use your actual unique table names. Apple’s reference help confirms cross-table TableName::Cell notation and standard B2:B6 ranges. [A4]
Advanced: compare a measured statistic against the critical t
For a two-sided decision at α=0.05 and df=10, =TINV(0.05,10) supplies 2.22813885198628. A signed observed t magnitude greater than that cutoff lies in the two tails (assuming the reference t model is appropriate). Conceptually:
=ABS(TestInputs::B2)>TINV(0.05,TestInputs::C2)
This returns the logical comparison of a statistic stored in TestInputs B2 against a t cutoff based on df in C2; it does not calculate a p-value or verify distributional assumptions. Do not swap the probability for 0.95 expecting a 95%-confidence cutoff: that would produce a very small positive t threshold. [M1, A4]
A useful distribution consistency check is:
=TDIST(TINV(0.05,10),10,2)
Mathematically the expected result is 0.05, within ordinary floating-point tolerance. Apple’s TDIST explicitly uses a two-tail mode of 2. Neither this check nor the statistical tests were executed natively in Numbers. [A2, M1]
Problems and fixes
| Symptom | Diagnosis and action |
|---|---|
| Probability rejected | Apple’s domain is 0 < p < 1. Do not use p=0, p=1, or percentages expressed as 5 instead of 0.05. [A1] |
| Unexpectedly tiny critical t | Ensure you passed α (the two-tail area) instead of confidence level 1−α. =TINV(0.95,df) does not give the conventional 95% two-sided critical value. [M1] |
| Incorrect degrees of freedom | Supply the appropriate df ≥ 1. In the one-sample example use n−1, not n; for other analyses, do not assume that rule. [A1, M2] |
| One-sided cutoff needed | The TINV parameterization evidenced by Apple’s example is two-tailed. For a one-sided α<0.5, the positive right-tail cutoff is TINV(2*α,df) mathematically; direction and hypotheses must be handled separately. [M1] |
| Formula separator error | If decimal comma is active, follow Apple’s semicolon guidance. Avoid changing both decimal and list separators arbitrarily. [A3] |
| Wrong references when filling | Use Numbers’ documented absolute/relative reference controls or explicit table names. [A4] |
| Too few observations for sample interval | With n≤1, df=n−1 falls below Apple’s documented minimum. More importantly, check statistical assumptions before reporting the interval. [A1, M2] |
Compatibility and version scope
- Apple function source: Formulas and Functions Help 15.4, checked 13 October 2026. The current English catalogue contains one canonical
TINVentry. Its date of original introduction in past Numbers releases is not established. [A1, A5] - System versions: Apple’s Numbers Mac release page describes Numbers 15.4 and macOS 26 as the minimum for that version. This is not evidence that TINV was newly added in 15.4. This guide has not tested formula execution on Mac, iPad, iPhone or web. [A6]
- Naming:
TINVis an exact uppercase English identifier; do not conflate it with inverse functions whose tail probability conventions differ. [A1, M1]
Related Apple references (destinations checked 2026-10-13)
- TINV — Apple syntax and worked example [A1]
- TDIST — one- versus two-tailed probability [A2]
- Functions overview — separators and required arguments [A3]
- Refer to cells in formulas — ranges and table names [A4]
Dated claim-level evidence
- [A1] Apple, “TINV,” Function Help 15.4, checked 2026-10-13:
https://support.apple.com/guide/functions/tinv-ffa5a023a7/15.4/web/1.0— exactly two required arguments; p strictly (0,1), df≥1; official example0.88, 2. - [A2] Apple, “TDIST,” Function Help 15.4, checked 2026-10-13:
https://support.apple.com/guide/functions/tdist-ffa5b41355/15.4/web/1.0— tail selector1or2, two published right/two-tail values, link to TINV. - [A3] Apple, “Functions overview,” checked 2026-10-13:
https://support.apple.com/guide/functions/functions-overview-ffa083bef0c/15.4/web/1.0— decimal-comma locale separators. - [A4] Apple, “Refer to cells in formulas,” checked 2026-10-13:
https://support.apple.com/en-au/guide/functions/ffa550d691/15.4/web/1.0— ranges and table-qualified references. - [A5] Apple, “List of functions by category,” checked 2026-10-13:
https://support.apple.com/en-au/guide/functions/ffa148e9cd0/web— listed canonical English identifier. - [A6] Apple, “What’s new in Numbers on Mac,” checked 2026-10-13:
https://support.apple.com/en-sa/guide/numbers/tan700f60676/mac— current version platform minimum and general statistical changes, not a function-by-function introduction list. - [M1] Independent SciPy: t inverse survival
scipy.stats.t.isf(p/2,df)andscipy.stats.t.sf(x,df), checked 2026-10-13. No native Numbers formula runs. - [M2] Independent Python: sample mean,
statistics.stdev, and t critical × sample standard error, checked 2026-10-13. No native Numbers formula runs.