T Score Calculator
The small-sample standardiser: t = (x̄ − μ)·√n / s with its degrees of freedom, the t-versus-z decision made explicit, and the p-value boundary drawn where the mathematics actually sits.
T Score Calculator
Results recalculate instantly on every keystroke. Nothing you type is transmitted.
What this result does not account for
- Statistic and df only — p-values are refused, not approximated
- One-sample form; paired and two-sample t are separate tools
In short: A sample of 16 measurements averages 52 against a claimed population mean of 50, with sample standard deviation 5: t = (52 − 50)·√16 / 5 = 1.6, on df = 16 − 1 = 15 degrees of freedom. The √n factor is the whole difference from a z score — the uncertainty of the SAMPLE mean shrinks with sample size, so the statistic divides by s/√n, not s. What the page refuses is the p-value: converting t = 1.6 into a tail probability needs the inverse of the t distribution’s curve, which this page does not implement — and a probability that cannot be computed is printed as a boundary, never a number that looks like one.
Formula
t = (x̄ − μ) / (s / √n) · df = n − 1
the t distribution is the z distribution with the population sigma replaced by an estimate — the extra wobble shows up as heavier tails, governed by df = n − 1.
Worked Example
- Standardise honestly. subtract the claimed mean, divide by the sample mean’s own uncertainty s/√n.
- Count the freedom. df = n − 1 — the same Bessel payment the sample standard deviation already made.
- Stop at the boundary. the statistic and its df are exact; the p-value needs an inverse-t this page does not ship, and the boundary is printed rather than papered over.
x̄ = 52, μ = 50, s = 5, n = 16: t = 1.6 on df = 15. At n = 121 the same gap gives t = 4.4 on df = 120 — larger samples make the same difference louder. At n = 4 it is t = 1.6 on df = 3 — small samples pay heavier tails for the same statistic.
Strengths & Limits Of This Model
Where this engine is strong
- The t-vs-z decision is stated per input, not assumed
- df computed and carried with the statistic
Where it stops
- No p-values by design
- No one-sample t interval (z-based CI lives next door)
Practical Use Cases
Lab studies
small-n comparisons against a reference value
Process pilots
a trial batch against the spec mean
Coursework
the statistic and df exactly, with the table lookup left where it belongs
Methodology & Editorial Standards
Computation runs in IEEE-754 double precision at full internal precision; rounding to two decimal places occurs strictly at the display layer, so no cumulative drift enters the result. All monetary outputs use accounting presentation — grouped thousands, two decimals, negatives in parentheses — so figures can be transcribed directly into a model or working paper. Division-by-zero and out-of-domain inputs return an em-dash rather than a misleading number.
This engine was reconciled against an independent reference implementation and hand-verified for the worked example above before release. Our full five-stage review process is published on the About Us page.
Disclaimer. This calculator is provided for informational and modelling purposes only and does not constitute financial, tax, legal, medical, or engineering advice. Verify all figures with a qualified professional before acting on them.
T Score Calculator — 8 Expert FAQs
8 analyst-written answers to the questions practitioners actually ask — optimised for voice and answer-engine retrieval.
Why does the formula divide by s/√n instead of s?
Because the question is about the SAMPLE MEAN, not individual values. Means of n values wobble less than single values — their spread shrinks by √n — so the honest standardiser divides by the mean’s own uncertainty. A z score answers “where is this value?”; a t statistic answers “where is this AVERAGE, given it was estimated?”.
What are degrees of freedom, really?
The number of independent pieces of information left after the estimation is done. Sixteen measurements give the sample mean for free, but the deviations from that mean are constrained to sum to zero — fifteen of them carry information, the sixteenth is determined. df = n − 1 is that subtraction, the same one Bessel’s correction makes in the sample standard deviation.
When should I use t instead of z?
Whenever the population standard deviation is unknown and was itself estimated from the sample — which in practice is almost always. The estimated sigma wobbles, and the t distribution’s heavier tails price that wobble in. As n grows the estimate stabilises and t collapses onto z; past a few hundred observations the difference is smaller than the display.
Why does this page refuse to print a p-value?
Because the p-value is the area under the t curve beyond the statistic, which needs the inverse of a parameterised special function. That machinery is not implemented here — and a page that cannot compute a number must not print one that looks like it did. The statistic and df are exact; carrying them to a printed table (or the confidence interval page for the z-based route) keeps every figure on this site computed or absent.
Is a bigger t always more significant?
For the same df, yes — the t distribution is fixed by df alone, so larger |t| always sits further into the tail. Across different df the comparison bends: t = 1.6 at df = 3 is far less surprising than t = 1.6 at df = 120, because heavier tails make extreme values cheaper. That is why the statistic is always quoted WITH its df.
What sample size makes t collapse into z?
There is no cliff — the tails thin continuously. By df ≈ 30 the difference is teaching-scale; by df ≈ 120 it is display-scale. The page says where your n sits on that slide rather than drawing an arbitrary line.
Can the sample mean gap be negative?
Yes — t carries the sign of x̄ − μ, and the sign says which side of the claim your sample landed on. Magnitude drives the comparison; the sign is direction.
What if n is 1?
Refused: with one value there is no sample standard deviation at all, and df = 0 has no distribution. The statistic needs at least two observations to exist.