Written and maintained by CASRAI Editorial Board
Published · Last updated
On this page
pt() in R returns the cumulative probability of the Student t distribution: given a t value and degrees of freedom, it gives P(T ≤ t). Researchers use it mainly to turn a t statistic into a p-value; its inverse, qt(), turns a probability into a critical value. This guide shows which of the four t-distribution functions to call for which task, with a worked example you can reproduce.
What pt() is used for in research
R’s stats package groups four functions for the Student t distribution on one manual page (“The Student t Distribution”): dt (density), pt (distribution function), qt (quantile function) and rt (random generation). Most analysis work needs only pt and qt: pt converts an observed t statistic into a tail probability, and qt converts a chosen tail probability (such as 0.975) into the t value that cuts it off. For background on the distribution itself, see what the t distribution is.
Access: nothing to install
All four functions are part of the stats package, which ships with base R. Download R from the R Project and the functions are available without loading any package.
Which function answers which question
| Question | Function | Call |
|---|---|---|
| Probability below a t value (lower tail) | pt | pt(q, df) |
| Probability above a t value (upper tail) | pt | pt(q, df, lower.tail = FALSE) |
| t value that cuts off a given lower-tail probability (critical value) | qt | qt(p, df) |
| Height of the density curve at a t value (for plotting) | dt | dt(x, df) |
| Random draws from a t distribution (simulation) | rt | rt(n, df) |
The full signatures on the manual page are dt(x, df, ncp, log = FALSE), pt(q, df, ncp, lower.tail = TRUE, log.p = FALSE), qt(p, df, ncp, lower.tail = TRUE, log.p = FALSE) and rt(n, df, ncp). The argument df is the degrees of freedom; ncp is an optional non-centrality parameter, and if it is omitted the central t distribution is used.
Step by step: how to solve for a p-value in R
This is a worked example with made-up numbers, not data from a real study. Suppose a one-sample t statistic is 2.31 with 14 degrees of freedom.
- Lower-tail probability.
pt(2.31, df = 14)returns about 0.9817, the probability of a t value at or below 2.31. - Upper-tail (one-sided) p-value.
pt(2.31, df = 14, lower.tail = FALSE)returns about 0.0183. Withlower.tail = FALSEthe probability is P(T > t) rather than P(T ≤ t). - Two-sided p-value.
2 * pt(-abs(2.31), df = 14)returns about 0.0366. Using-abs(t)puts the value in the lower tail whatever the sign of the statistic, then doubles it for the two tails. Forgettingabs()with a positive t gives a p-value above 1. - Critical value.
qt(0.975, df = 14)returns about 2.1448, the cutoff for a 95% two-sided test. Because 2.31 exceeds it, the two-sided p-value falls below 0.05, which agrees with step 3.
For a left-tailed or two-tailed hypothesis, see left-tailed tests and two-tailed hypothesis testing. The same pattern with the normal distribution is covered in pnorm in R; the t distribution approaches it as the degrees of freedom grow.
qt vs pt: inverse functions
qt is the inverse of pt. pt goes from a t value to a probability; qt goes from a probability to a t value. Round trip: pt(qt(0.975, 14), 14) returns 0.975. Use pt when you have a statistic and want a p-value, and qt when you have a significance level and want the critical value or a confidence-interval multiplier.
qt in R
qt(p, df) is the quantile function. Its lower.tail argument works as for pt: qt(0.05, df = 14, lower.tail = FALSE) equals qt(0.95, df = 14), about 1.7613. The R manual’s own example, qt(.975, df = c(1:10, 20, 50, 100, 1000)), shows the 97.5% quantile shrinking toward 1.96 as df grows (about 2.228 at df = 10, 2.086 at 20, 2.009 at 50, 1.984 at 100 and 1.962 at 1000).
rt in R
rt(n, df) draws n random values from a t distribution, for example rt(1000, df = 14) for a simulation. The result has length n. Use set.seed() first if you need reproducible draws.
dt in R
dt(x, df) gives the density, useful for drawing the curve, for example with curve(dt(x, df = 14), from = -4, to = 4).
Reading the output
- All four functions return numbers, not test reports.
ptandqtare vectorised: arguments are recycled to the length of the longest, sopt(c(1, 2, 3), df = 14)returns three probabilities. lower.tail = TRUE(the default) gives P(X ≤ x);FALSEgives P(X > x).log.p = TRUEreturns log probabilities, useful for very small tails.- The central t distribution has mean 0 for df > 1 and variance df/(df − 2) for df > 2.
- If you run a test with
t.test(), the returned object already containsstatistic,parameter(the degrees of freedom),p.valueandconf.int, so you can compare your manualptcalculation tot.test(x)$p.value. See t-test in R. By defaultt.test()uses the Welch test (var.equal = FALSE) for two samples, so its degrees of freedom may not be an integer;ptandqtaccept non-integer df.
Troubleshooting
- NaN with a warning: invalid arguments, such as non-positive
df, return NaN with a warning.dfmust be positive; it may be non-integer, andInfis allowed. - p-value above 1: usually
2 * pt(t, df)with a positive t. Use2 * pt(-abs(t), df). - p-value near 1 when you expected near 0: you probably need
lower.tail = FALSE(upper tail), or the opposite sign. - Setting
ncp = 0is not the same as omitting it: the manual notes thatncp = 0uses the non-central algorithm, which differs from the central one. Leavencpout for ordinary tests; the non-central form is mainly for power calculations, and results are documented as accurate only for |ncp| up to 37.62.
What are the t and F distributions?
The Student t distribution is a symmetric, bell-shaped distribution centred on 0 that is used for test statistics when the population standard deviation is estimated from the sample; its single parameter is the degrees of freedom. The F distribution is a different distribution, used for ratios of variances such as in ANOVA, and it has two degrees-of-freedom parameters. R follows the same d/p/q/r naming for it (df, pf, qf, rf), and for the normal distribution (dnorm, pnorm, qnorm, rnorm).
What does the t() function do in R?
Nothing related to the t distribution. In base R, t(x) is the matrix transpose: given a matrix or data frame, it returns the transpose (R base documentation, “Matrix Transpose”). The t distribution functions are dt, pt, qt and rt.
Free alternatives
| Task | R | Python (SciPy) |
|---|---|---|
| CDF (lower tail) | pt(t, df) |
scipy.stats.t.cdf(t, df) |
| Upper tail | pt(t, df, lower.tail = FALSE) |
scipy.stats.t.sf(t, df) |
| Quantile | qt(p, df) |
scipy.stats.t.ppf(p, df) |
The SciPy results for the worked example above (cdf 0.9817, sf 0.0183, ppf 2.1448) match the R values given.
FAQ
What does pt give in R?
The cumulative probability of the Student t distribution: P(T ≤ q) for a t value q and degrees of freedom df, or P(T > q) with lower.tail = FALSE.
What is pt in R code?
It is the distribution-function member of the t-distribution family in the stats package: pt(q, df, ncp, lower.tail = TRUE, log.p = FALSE).
How do I solve for a p-value in R?
Two-sided: 2 * pt(-abs(t), df). One-sided upper: pt(t, df, lower.tail = FALSE). See the worked example above and the p-value definition.
What does the t() function do in R?
It transposes a matrix or data frame; it is unrelated to the t distribution.
What are the t and F distributions?
Both are probability distributions used for test statistics: t has one degrees-of-freedom parameter, F has two. R provides dt/pt/qt/rt and df/pf/qf/rf.
What is the difference between qt and pt?
pt maps a t value to a probability; qt maps a probability to a t value.
Sources
- R Core Team, “The Student t Distribution,” R stats manual: stat.ethz.ch/R-manual/R-devel/library/stats/html/TDist.html (checked 2026-11-07).
- R Core Team, “The Normal Distribution” (lower.tail wording): Normal.html (checked 2026-11-07).
- R Core Team, “Student’s t-Test” (t.test): t.test.html (checked 2026-11-07).
- SciPy documentation, “scipy.stats.t”: docs.scipy.org/doc/scipy/reference/generated/scipy.stats.t.html (checked 2026-11-07).
- R Core Team, “Matrix Transpose”: base/html/t.html (checked 2026-11-07).







