Student's t-distribution explained

Student's
Type:	density
Parameters:	\nu > 0 degrees of freedom (real, almost always a positive integer)
Support:	x \in ( - inf ty, inf ty)
Pdf:	style (\Gamma \left( (\nu+1)/(2) \right))/(\sqrt{\pi \nu) \Gamma \left( (\nu)/(2) \right)} \left( 1 + (~ x ^{2})/(\nu) \right) ^{- (\nu +1)/(2)}
Cdf:	$$\backslash begin\; \backslash \; \backslash frac\; +\; x\backslash \; \backslash Gamma\; \backslash left(\backslash frac\; \backslash right)\; \backslash times\; \backslash \backslash [0.5em]\; \backslash frac\; \backslash ,\; \backslash end$$ where {} _{2} F _{1( , ; ; )} is the hypergeometric function
Mean:	0 for \nu > 1 , otherwise undefined
Median:	0
Mode:	0
Variance:	style (\nu)/(\nu-2) for \nu > 2 , for 1 < \nu \le 2 , otherwise undefined
Skewness:	0 for \nu > 3 , otherwise undefined
Kurtosis:	style (6)/(\nu-4) for \nu > 4 , ∞ for 2 < \nu \le 4 , otherwise undefined
Entropy:	\begin{matrix} (\nu +1)/(2) \left[ \psi \left( (\nu +1)/(2) \right) - \psi \left( (\nu)/(2) \right) \right] \\[ 0 . 5 em] + ln \left[ \sqrt{\nu } {B}\left( (\nu)/(, 2) (1)/(2) \right)\right] {\scriptstyle (nats) }\ \end where\psi( ) is the digamma function,{ B}( , ) is the beta function.
Mgf:	undefined
Char:	style (\left( \sqrt{\nu)/(|t| \right)) ^{\nu/ 2} K _{\nu/2} \left( \sqrt{\nu } |t| \right) }{ \Gamma(\nu/ 2 ) 2 ^{\nu/2} } for \nu > 0 K _{\nu(x)} is the modified Bessel function of the second kind^[1]
Es:	\mu + s \left( (\nu + T ^{-1} ( 1- p) ^{2 x \tau\left(} T ^{-1} ( 1- p) ^{2} \right))/((\nu -1 )( 1- p)) \right) Where T ^{-1} ( ) is the inverse standardized Student  CDF, and \tau( ) is the standardized Student t PDF.^[2]

In probability and statistics, Student's  distribution (or simply the  distribution)

t _{\nu} isa continuous probability distribution that generalizes the standard normal distribution. Like the latter, it is symmetric around zero and bell-shaped.

However,

t _{\nu} has heavier tails and the amount of probability mass in the tails is controlled by the parameter \nu ~ ~. For \nu = 1 the Student's distribution t _{\nu} becomes the standard Cauchy distribution, which has very "fat" tails; whereas for \nu → inf ty it becomes the standard normal distribution l{N}(0,1) which has very "thin" tails.

The Student's  distribution plays a role in a number of widely used statistical analyses, including Student's  test for assessing the statistical significance of the difference between two sample means, the construction of confidence intervals for the difference between two population means, and in linear regression analysis.

In the form of the location-scale  distribution

lst(\mu, \tau ^{2,} \nu) it generalizes the normal distribution and also arises in the Bayesian analysis of data from a normal family as a compound distribution when marginalizing over the variance parameter.

History and etymology

In statistics, the  distribution was first derived as a posterior distribution in 1876 by Helmert^{[3] [4] [5]} and Lüroth.^{[6] [7] [8]} As such, Student's t-distribution is an example of Stigler's Law of Eponymy. The  distribution also appeared in a more general form as Pearson type IV distribution in Karl Pearson's 1895 paper.^[9]

In the English-language literature, the distribution takes its name from William Sealy Gosset's 1908 paper in Biometrika under the pseudonym "Student".^[10] One version of the origin of the pseudonym is that Gosset's employer preferred staff to use pen names when publishing scientific papers instead of their real name, so he used the name "Student" to hide his identity. Another version is that Guinness did not want their competitors to know that they were using the  test to determine the quality of raw material.^{[11] [12]}

Gosset worked at the Guinness Brewery in Dublin, Ireland, and was interested in the problems of small samples – for example, the chemical properties of barley where sample sizes might be as few as 3. Gosset's paper refers to the distribution as the "frequency distribution of standard deviations of samples drawn from a normal population". It became well known through the work of Ronald Fisher, who called the distribution "Student's distribution" and represented the test value with the letter .^{[13] [14]}

Definition

Probability density function

Student's  distribution has the probability density function (PDF) given by

f(t) = (\Gamma\left( (\nu +1)/(\right) 2))/(\sqrt{\pi \nu) \Gamma\left( (\nu)/(2) \right)} \left( 1 + (~ t ^{2})/(\nu) \right) ^{-(\nu+1)/2} ,

where

\nu is the number of degrees of freedom and \Gamma is the gamma function. This may also be written asf(t) = (1)/({ B}\left( \sqrt{\nu) (1)/(, 2) (\nu)/(2) \right) } \left( 1 + (t ^{2})/(\nu) \right) ^{-(\nu+1)/2} ,

where

{ B} is the Beta function. In particular for integer valued degrees of freedom \nu we have:

For

\nu > 1 and even,(\Gamma\left( (\nu +1)/(2) \right))/(\sqrt{\pi \nu) \Gamma\left( (\nu)/(\right) } = 2) (1)/(} ⋅ 2 \sqrt{\nu) ((\nu -1 ) ⋅ (\nu -3 ) … 5 ⋅ 3)/((\nu -2 ) ⋅ (\nu -4 ) … 4 ⋅ 2) ~.

For

\nu > 1 and odd,(\Gamma\left( (\nu +1)/(2) \right))/(\sqrt{\pi \nu) \Gamma\left( (\nu)/(2) \right)} = (1)/(} ⋅ \pi \sqrt{\nu) ((\nu -1 ) ⋅ (\nu -3 ) … 4 ⋅ 2)/((\nu -2 ) ⋅ (\nu -4 ) … 5 ⋅ 3) ~.

The probability density function is symmetric, and its overall shape resembles the bell shape of a normally distributed variable with mean 0 and variance 1, except that it is a bit lower and wider. As the number of degrees of freedom grows, the  distribution approaches the normal distribution with mean 0 and variance 1. For this reason

{ \nu } is also known as the normality parameter.^[15]

The following images show the density of the  distribution for increasing values of

\nu ~. The normal distribution is shown as a blue line for comparison. Note that the  distribution (red line) becomes closer to the normal distribution as \nu increases.

Cumulative distribution function

The cumulative distribution function (CDF) can be written in terms of, the regularizedincomplete beta function. For

F(t) = \int ^{t f(u) \operatorname{d}u} _{-inf ty} ~ = ~ 1 - (1)/(2) I _{x(t)} \left( (\nu)/(, 2) (1)/(2) \right) ,

where

x(t) = (\nu)/(t ^{2+\nu}) ~.

Other values would be obtained by symmetry. An alternative formula, valid for

t ^{2} < \nu , is\int ^{t} _{-inf ty} f(u) \operatorname{d}u ~ = ~ (1)/(2) + t (\Gamma\left( (\nu +1)/(2) \right))/(\sqrt{\pi \nu) \Gamma\left( (\nu)/(2) \right) } {} _{2} F _{1 \left( (1)/(2)} , (\nu +1)/(2) ; (3)/(; - 2) (~ t ^{2})/(\nu) \right) ,

where

{} _{2} F _{1( , ; ; )} is a particular instance of the hypergeometric function.

For information on its inverse cumulative distribution function, see .

Special cases

Certain values of

\nu give a simple form for Student's t-distribution.

\nu	PDF	CDF	notes
1	(1)/(\pi ( 1+ t ^{2)})	(1)/(2) + (1)/(\pi) \arctan( t )	See Cauchy distribution
2	(1)/(\left( 1+ 2 \sqrt{ 2) (t ^{2})/(2) \right) ^{3/2} }\	(1)/(+ 2) (t)/(2 \sqrt{ 2) \sqrt{ 1 + (~ t ^{2})/(2) } }
3	(2)/(\pi \sqrt{ 3) \left( 1 + (~ t ^{2})/(3) \right) ^{2 }}	(1)/(2) + (1)/(\pi) \left[ (\left( (t)/(\sqrt{ 3))/(\right)) }{ \left( 1 + (~ t ^{2})/(3) \right) } + \arctan\left( (t)/(\sqrt{ 3) } \right) \right]
4	(3)/(8 \left( 1 + (~ t ^{2})/(4) \right) ^{5/2})	(1)/(2) + (3)/(\left[ 8) (t)/(\sqrt{ 1 + (~ t ^{2})/(4) ~) } \right] \left[ 1 - (~ t ^{2})/(12 \left( 1 + (~ t ^{2})/(4) \right)) \right]
5	(8)/(\left( 1+ 3 \pi \sqrt{ 5) (t ^{2})/(5) \right) ^{3 }}	(1)/(2) + (1)/(\pi) { \left[ (t)/(\sqrt{ 5) \left( 1 + (t ^{2})/(5) \right) } \left( 1 + (2)/(3 \left( 1 + (t ^{2})/(5) \right)) \right) + \arctan\left( (t)/(\sqrt{ 5) } \right)\right]}
inf ty	(1)/(\sqrt{ 2 \pi) } e ^{- t ^{2/2}}	(1)/(2) {\left[ 1 + \operatorname{erf}\left( (t)/(\sqrt{ 2) } \right) \right]}	See Normal distribution, Error function

Moments

For

\nu > 1 , the raw moments of the  distribution are\operatorname{ E}\left\{ T ^{k \right\}} = \begin{cases} 0 & k odd , 0 < k < \nu , \ {} \\ (1)/(\Gamma\left( \sqrt{\pi) (\nu)/(\right)} \left[ \Gamma\left( 2) (k +1)/(\right) \Gamma\left( 2) (\nu - k)/(2) \right) \nu ^{(k)/(2)} \right] & k even , 0 < k < \nu ~.\\ \end{cases}

Moments of order

\nu or higher do not exist.^[16]

The term for

0 < k < \nu , even, may be simplified using the properties of the gamma function to\operatorname{ E}\left\{ T ^{k \right\}} = \nu ^{(k)/(2)} \prod ^{k/ 2} _{j =1} (~ 2 j -1 ~)/(\nu -2 j) k even , 0 < k < \nu ~.

For a  distribution with

\nu degrees of freedom, the expected value is 0 if \nu > 1 , and its variance is (\nu)/(\nu -2) if \nu > 2 ~. The skewness is 0 if \nu > 3 and the excess kurtosis is (6)/(\nu -4) if \nu > 4 ~.

Location-scale  distribution

Location-scale transformation

Student's  distribution generalizes to the three parameter location-scale  distribution

l{lst}(\mu, \tau ^{2, \nu)} by introducing a location parameter \mu and a scale parameter \tau ~. WithT \sim t _{\nu} and location-scale family transformationX = \mu + \tau T we getX \sim l{lst}(\mu, \tau ^{2, \nu)} ~.

The resulting distribution is also called the non-standardized Student's  distribution.

Density and first two moments

The location-scale distribution has a density defined by:^[17]

p(x\mid \nu,\mu,\tau) = (\Gamma\left( (\nu +1)/(\right) 2))/(\Gamma\left( (\nu)/(2) \right) \sqrt{\pi \nu) \tau } \left( 1 + (1)/(\left( \nu) (x - \mu)/(\tau) \right) ^{2 \right)} ^{-(\nu+1)/2}

Equivalently, the density can be written in terms of

\tau ^{2} :p(x \mid \nu, \mu, \tau ^{2)} = (\Gamma( (\nu +1)/(2) ))/(\Gamma\left( (\nu)/(\right) \sqrt{\pi \nu \tau ^{2} 2)) } \left( 1 + (1)/(\nu) ((x - \mu) ^{2})/(\tau ^{2}) \right) ^{-(\nu+1)/2}

Other properties of this version of the distribution are:^[17]

\begin{align} \operatorname{ E}\{ X \} & = \mu & for \nu > 1 ,\\ \operatorname{var}\{ X \} & = \tau ^{2 (\nu)/(\nu -2)} & for \nu > 2 ,\\ \operatorname{mode}\{ X \} & = \mu ~. \end{align}

Special cases

• If

X follows a location-scale  distribution X \sim l{lst}\left(\mu, \tau ^{2, \nu\right)} then for \nu → inf ty X is normally distributed X \sim N \left(\mu, \tau ^{2\right)} with mean \mu and variance \tau ^{2} ~.

• The location-scale  distribution

l{lst}\left(\mu, \tau ^{2, \nu=1} \right) with degree of freedom \nu =1 is equivalent to the Cauchy distribution Cau \left(\mu, \tau\right) ~.

• The location-scale  distribution

l{lst}\left(\mu=0, \tau ^{2=1, \nu\right)} with \mu =0 and \tau ^{2=1} reduces to the Student's  distribution t _{\nu} ~.

How the  distribution arises (characterization)

As the distribution of a test statistic

Student's t-distribution with

\nu degrees of freedom can be defined as the distribution of the random variable T with^{[18] [19]} T = (Z)/(\sqrt{V/\nu) } = Z \sqrt,

where

• Z is a standard normal with expected value 0 and variance 1;
• V has a chi-squared distribution with

\nu degrees of freedom;

• Z and V are independent;

A different distribution is defined as that of the random variable defined, for a given constant μ, by

(Z + \mu)\sqrt{ (\nu)/(V) }.This random variable has a noncentral t-distribution with noncentrality parameter μ. This distribution is important in studies of the power of Student's t-test.

Derivation

Suppose X₁, ..., X_n are independent realizations of the normally-distributed, random variable X, which has an expected value μ and variance σ². Let

\overline{X} _{n} = (1)/(n) (X _{1+ … +X} _{n)}

be the sample mean, and

S ^{2} _{n} = (1)/(n -1) \sum ^{n} _{i =1} \left(X _{i} - \overline{X} ^{2} _{n\right)}

be an unbiased estimate of the variance from the sample. It can be shown that the random variable

V = (n -1 ) (S ^{2} _{n})/(\sigma ^{2})

has a chi-squared distribution with

\nu = n - 1 degrees of freedom (by Cochran's theorem).^[20] It is readily shown that the quantityZ = \left(\overline{X} _{n} - \mu\right) (\sqrt{n)/() }

is normally distributed with mean 0 and variance 1, since the sample mean

\overline{X} _{n} is normally distributed with mean μ and variance σ²/n. Moreover, it is possible to show that these two random variables (the normally distributed one Z and the chi-squared-distributed one V) are independent. Consequently the pivotal quantity

$T\; \backslash equiv\; \backslash frac\; =\; \backslash left(\backslash overline\_n\; -\; \backslash mu\backslash right)\; \backslash frac,$

which differs from Z in that the exact standard deviation σ is replaced by the random variable S_n, has a Student's t-distribution as defined above. Notice that the unknown population variance σ² does not appear in T, since it was in both the numerator and the denominator, so it canceled. Gosset intuitively obtained the probability density function stated above, with

\nu equal to n − 1, and Fisher proved it in 1925.^[13]

The distribution of the test statistic T depends on

\nu , but not μ or σ; the lack of dependence on μ and σ is what makes the t-distribution important in both theory and practice.

Sampling distribution of t-statistic

The  distribution arises as the sampling distributionof the  statistic. Below the one-sample  statistic is discussed, for the corresponding two-sample  statistic see Student's t-test.

Unbiased variance estimate

Let

x _{1,} \ldots, x _{n} \sim { l N}(\mu, \sigma ^{2)} be independent and identically distributed samples from a normal distribution with mean \mu and variance \sigma ^{2} ~. The sample mean and unbiased sample variance are given by:\begin{align} \bar{x} & = (x _{1+ … +x} _{n})/(n) , \\[ 5 pt] s ^{2} & = (1)/(n -1) \sum ^{n} _{i =1} (x _{i} - \bar{x}) ^{2} ~. \end{align}

The resulting (one sample)  statistic is given by

t = (\bar{x)/(-) \mu}{ \sqrt{s ^{2/n } }} \sim t _{n-1} ~.

and is distributed according to a Student's  distribution with

n - 1 degrees of freedom.

Thus for inference purposes the  statistic is a useful "pivotal quantity" in the case when the mean and variance

(\mu, \sigma ^{2)} are unknown population parameters, in the sense that the  statistic has then a probability distribution that depends on neither \mu nor \sigma ^{2} ~.

ML variance estimate

Instead of the unbiased estimate

s ^{2} we may also use the maximum likelihood estimates ^{2} _{ML} = (1)/(n) \sum ^{n} _{i =1} (x _{i} - \bar{x}) ^{2} yielding the statistict _{ML} = (\bar{x)/(-) \mu}{\sqrt{s ^{2} _{ML /n }}} = \sqrt{ (n)/(n -1) } t ~. This is distributed according to the location-scale  distribution:t _{ML} \sim l{lst}(0, \tau ^{2=n/(n-1), n-1)} ~.

Compound distribution of normal with inverse gamma distribution

\mu and unknown variance, with an inverse gamma distribution placed over the variance with parameters a = (\nu)/(2) and b = (\nu \tau ^{2})/(2) ~. In other words, the random variable X is assumed to have a Gaussian distribution with an unknown variance distributed as inverse gamma, and then the variance is marginalized out (integrated out).

Equivalently, this distribution results from compounding a Gaussian distribution with a scaled-inverse-chi-squared distribution with parameters

\nu and \tau ^{2} ~. The scaled-inverse-chi-squared distribution is exactly the same distribution as the inverse gamma distribution, but with a different parameterization, i.e. \nu = 2 a, {\tau} ^{2} = (b)/(a) ~.

The reason for the usefulness of this characterization is that in Bayesian statistics the inverse gamma distribution is the conjugate prior distribution of the variance of a Gaussian distribution. As a result, the location-scale  distribution arises naturally in many Bayesian inference problems.^[21]

Maximum entropy distribution

Student's  distribution is the maximum entropy probability distribution for a random variate X for which

\operatorname{ E}\left\{ ln(\nu+X ^{2) \right\}} is fixed.^[22]

Further properties

Monte Carlo sampling

There are various approaches to constructing random samples from the Student's  distribution. The matter depends on whether the samples are required on a stand-alone basis, or are to be constructed by application of a quantile function to uniform samples; e.g., in the multi-dimensional applications basis of copula-dependency. In the case of stand-alone sampling, an extension of the Box–Muller method and its polar form is easily deployed.^[23] It has the merit that it applies equally well to all real positive degrees of freedom,, while many other candidate methods fail if is close to zero.^[23]

Integral of Student's probability density function and -value

The function is the integral of Student's probability density function, between   and, for It thus gives the probability that a value of t less than that calculated from observed data would occur by chance. Therefore, the function can be used when testing whether the difference between the means of two sets of data is statistically significant, by calculating the corresponding value of and the probability of its occurrence if the two sets of data were drawn from the same population. This is used in a variety of situations, particularly in  tests. For the statistic, with degrees of freedom, is the probability that would be less than the observed value if the two means were the same (provided that the smaller mean is subtracted from the larger, so that It can be easily calculated from the cumulative distribution function of the  distribution:

A( t \mid \nu) = F _{\nu(t)} - F _{\nu(-t)} = 1 - I _{(\nu)/(\nu + t ^{2})} \left( (\nu)/(, 2) (1)/(2) \right),

where is the regularized incomplete beta function.

For statistical hypothesis testing this function is used to construct the p-value.

Related distributions

• The noncentral  distribution generalizes the  distribution to include a noncentrality parameter. Unlike the nonstandardized  distributions, the noncentral distributions are not symmetric (the median is not the same as the mode).
• The discrete Student's  distribution is defined by its probability mass function at r being proportional to:^[24] $$\backslash prod\_^k\; \backslash frac\; \backslash quad\; \backslash quad\; r=\backslash ldots,\; -1,\; 0,\; 1,\; \backslash ldots\; ~.$$ Here a, b, and k are parameters. This distribution arises from the construction of a system of discrete distributions similar to that of the Pearson distributions for continuous distributions.^[25]
• One can generate Student samples by taking the ratio of variables from the normal distribution and the square-root of the . If we use instead of the normal distribution, e.g., the Irwin–Hall distribution, we obtain over-all a symmetric 4 parameter distribution, which includes the normal, the uniform, the triangular, the Student  and the Cauchy distribution. This is also more flexible than some other symmetric generalizations of the normal distribution.
•  distribution is an instance of ratio distributions.

Uses

In frequentist statistical inference

Student's  distribution arises in a variety of statistical estimation problems where the goal is to estimate an unknown parameter, such as a mean value, in a setting where the data are observed with additive errors. If (as in nearly all practical statistical work) the population standard deviation of these errors is unknown and has to be estimated from the data, the  distribution is often used to account for the extra uncertainty that results from this estimation. In most such problems, if the standard deviation of the errors were known, a normal distribution would be used instead of the  distribution.

Confidence intervals and hypothesis tests are two statistical procedures in which the quantiles of the sampling distribution of a particular statistic (e.g. the standard score) are required. In any situation where this statistic is a linear function of the data, divided by the usual estimate of the standard deviation, the resulting quantity can be rescaled and centered to follow Student's  distribution. Statistical analyses involving means, weighted means, and regression coefficients all lead to statistics having this form.

Quite often, textbook problems will treat the population standard deviation as if it were known and thereby avoid the need to use the Student's  distribution. These problems are generally of two kinds: (1) those in which the sample size is so large that one may treat a data-based estimate of the variance as if it were certain, and (2) those that illustrate mathematical reasoning, in which the problem of estimating the standard deviation is temporarily ignored because that is not the point that the author or instructor is then explaining.

Hypothesis testing

A number of statistics can be shown to have  distributions for samples of moderate size under null hypotheses that are of interest, so that the  distribution forms the basis for significance tests. For example, the distribution of Spearman's rank correlation coefficient, in the null case (zero correlation) is well approximated by the distribution for sample sizes above about 20.

Confidence intervals

Suppose the number A is so chosen that

\operatorname{ P}\left\{ -A < T < A \right\} = 0 . 9 ,

when has a  distribution with degrees of freedom. By symmetry, this is the same as saying that satisfies

\operatorname{ P}\left\{ T < A \right\} = 0 . 95 ,

so A is the "95th percentile" of this probability distribution, or

A = t _{(0.05,n-1)} ~. Then\operatorname{ P}\left\{ -A < (\overline{X)/(_{n}) - \mu }{ S _{n/\sqrt{n }} } < A \right\} = 0 . 9 ,

and this is equivalent to

\operatorname{ P}\left\{ \overline{X} _{n} - A (S _{n})/(\sqrt{n) } < \mu < \overline{X} _{n} + A (S _{n})/(\sqrt{n) } \right\} = 0 . 9 .

Therefore, the interval whose endpoints are

\overline{X} _{n \pm} A (S _{n})/(\sqrt{n) }

is a 90% confidence interval for μ. Therefore, if we find the mean of a set of observations that we can reasonably expect to have a normal distribution, we can use the  distribution to examine whether the confidence limits on that mean include some theoretically predicted value – such as the value predicted on a null hypothesis.

It is this result that is used in the Student's  tests: since the difference between the means of samples from two normal distributions is itself distributed normally, the  distribution can be used to examine whether that difference can reasonably be supposed to be zero.

If the data are normally distributed, the one-sided confidence limit (UCL) of the mean, can be calculated using the following equation:

UCL _{1-\alpha} = \overline{X} _{n} + t _{\alpha,n-1} (S _{n})/(\sqrt{n) } ~. The resulting UCL will be the greatest average value that will occur for a given confidence interval and population size. In other words, \overline{X} _{n} being the mean of the set of observations, the probability that the mean of the distribution is inferior to is equal to the confidence

Prediction intervals

The  distribution can be used to construct a prediction interval for an unobserved sample from a normal distribution with unknown mean and variance.

In Bayesian statistics

The Student's  distribution, especially in its three-parameter (location-scale) version, arises frequently in Bayesian statistics as a result of its connection with the normal distribution. Whenever the variance of a normally distributed random variable is unknown and a conjugate prior placed over it that follows an inverse gamma distribution, the resulting marginal distribution of the variable will follow a Student's  distribution. Equivalent constructions with the same results involve a conjugate scaled-inverse-chi-squared distribution over the variance, or a conjugate gamma distribution over the precision. If an improper prior proportional to is placed over the variance, the  distribution also arises. This is the case regardless of whether the mean of the normally distributed variable is known, is unknown distributed according to a conjugate normally distributed prior, or is unknown distributed according to an improper constant prior.

Related situations that also produce a  distribution are:

• The marginal posterior distribution of the unknown mean of a normally distributed variable, with unknown prior mean and variance following the above model.
• The prior predictive distribution and posterior predictive distribution of a new normally distributed data point when a series of independent identically distributed normally distributed data points have been observed, with prior mean and variance as in the above model.

Robust parametric modeling

The  distribution is often used as an alternative to the normal distribution as a model for data, which often has heavier tails than the normal distribution allows for; see e.g. Lange et al.^[26] The classical approach was to identify outliers (e.g., using Grubbs's test) and exclude or downweight them in some way. However, it is not always easy to identify outliers (especially in high dimensions), and the  distribution is a natural choice of model for such data and provides a parametric approach to robust statistics.

A Bayesian account can be found in Gelman et al.^[27] The degrees of freedom parameter controls the kurtosis of the distribution and is correlated with the scale parameter. The likelihood can have multiple local maxima and, as such, it is often necessary to fix the degrees of freedom at a fairly low value and estimate the other parameters taking this as given. Some authors report that values between 3 and 9 are often good choices. Venables and Ripley suggest that a value of 5 is often a good choice.

Student's  process

For practical regression and prediction needs, Student's  processes were introduced, that are generalisations of the Student  distributions for functions. A Student's  process is constructed from the Student  distributions like a Gaussian process is constructed from the Gaussian distributions. For a Gaussian process, all sets of values have a multidimensional Gaussian distribution. Analogously,

X(t) is a Student  process on an interval I = [a,b] if the correspondent values of the process X(t _{1), \ldots ,} X(t _{n)} (t _{i} \in I ) have a joint multivariate Student  distribution.^[28] These processes are used for regression, prediction, Bayesian optimization and related problems. For multivariate regression and multi-output prediction, the multivariate Student  processes are introduced and used.^[29]

Table of selected values

The following table lists values for  distributions with degrees of freedom for a range of one-sided or two-sided critical regions. The first column is, the percentages along the top are confidence levels

\alpha , and the numbers in the body of the table are the t _{\alpha,n-1} factors described in the section on confidence intervals.

The last row with infinite gives critical points for a normal distribution since a  distribution with infinitely many degrees of freedom is a normal distribution. (See Related distributions above).

One-sided	75%	80%	85%	90%	95%	97.5%	99%	99.5%	99.75%	99.9%	99.95%
Two-sided	50%	60%	70%	80%	90%	95%	98%	99%	99.5%	99.8%	99.9%
1	1.000	1.376	1.963	3.078	6.314	12.706	31.821	63.657	127.321	318.309	636.619
2	0.816	1.061	1.386	1.886	2.920	4.303	6.965	9.925	14.089	22.327	31.599
3	0.765	0.978	1.250	1.638	2.353	3.182	4.541	5.841	7.453	10.215	12.924
4	0.741	0.941	1.190	1.533	2.132	2.776	3.747	4.604	5.598	7.173	8.610
5	0.727	0.920	1.156	1.476	2.015	2.571	3.365	4.032	4.773	5.893	6.869
6	0.718	0.906	1.134	1.440	1.943	2.447	3.143	3.707	4.317	5.208	5.959
7	0.711	0.896	1.119	1.415	1.895	2.365	2.998	3.499	4.029	4.785	5.408
8	0.706	0.889	1.108	1.397	1.860	2.306	2.896	3.355	3.833	4.501	5.041
9	0.703	0.883	1.100	1.383	1.833	2.262	2.821	3.250	3.690	4.297	4.781
10	0.700	0.879	1.093	1.372	1.812	2.228	2.764	3.169	3.581	4.144	4.587
11	0.697	0.876	1.088	1.363	1.796	2.201	2.718	3.106	3.497	4.025	4.437
12	0.695	0.873	1.083	1.356	1.782	2.179	2.681	3.055	3.428	3.930	4.318
13	0.694	0.870	1.079	1.350	1.771	2.160	2.650	3.012	3.372	3.852	4.221
14	0.692	0.868	1.076	1.345	1.761	2.145	2.624	2.977	3.326	3.787	4.140
15	0.691	0.866	1.074	1.341	1.753	2.131	2.602	2.947	3.286	3.733	4.073
16	0.690	0.865	1.071	1.337	1.746	2.120	2.583	2.921	3.252	3.686	4.015
17	0.689	0.863	1.069	1.333	1.740	2.110	2.567	2.898	3.222	3.646	3.965
18	0.688	0.862	1.067	1.330	1.734	2.101	2.552	2.878	3.197	3.610	3.922
19	0.688	0.861	1.066	1.328	1.729	2.093	2.539	2.861	3.174	3.579	3.883
20	0.687	0.860	1.064	1.325	1.725	2.086	2.528	2.845	3.153	3.552	3.850
21	0.686	0.859	1.063	1.323	1.721	2.080	2.518	2.831	3.135	3.527	3.819
22	0.686	0.858	1.061	1.321	1.717	2.074	2.508	2.819	3.119	3.505	3.792
23	0.685	0.858	1.060	1.319	1.714	2.069	2.500	2.807	3.104	3.485	3.767
24	0.685	0.857	1.059	1.318	1.711	2.064	2.492	2.797	3.091	3.467	3.745
25	0.684	0.856	1.058	1.316	1.708	2.060	2.485	2.787	3.078	3.450	3.725
26	0.684	0.856	1.058	1.315	1.706	2.056	2.479	2.779	3.067	3.435	3.707
27	0.684	0.855	1.057	1.314	1.703	2.052	2.473	2.771	3.057	3.421	3.690
28	0.683	0.855	1.056	1.313	1.701	2.048	2.467	2.763	3.047	3.408	3.674
29	0.683	0.854	1.055	1.311	1.699	2.045	2.462	2.756	3.038	3.396	3.659
30	0.683	0.854	1.055	1.310	1.697	2.042	2.457	2.750	3.030	3.385	3.646
40	0.681	0.851	1.050	1.303	1.684	2.021	2.423	2.704	2.971	3.307	3.551
50	0.679	0.849	1.047	1.299	1.676	2.009	2.403	2.678	2.937	3.261	3.496
60	0.679	0.848	1.045	1.296	1.671	2.000	2.390	2.660	2.915	3.232	3.460
80	0.678	0.846	1.043	1.292	1.664	1.990	2.374	2.639	2.887	3.195	3.416
100	0.677	0.845	1.042	1.290	1.660	1.984	2.364	2.626	2.871	3.174	3.390
120	0.677	0.845	1.041	1.289	1.658	1.980	2.358	2.617	2.860	3.160	3.373
∞	0.674	0.842	1.036	1.282	1.645	1.960	2.326	2.576	2.807	3.090	3.291
One-sided	75%	80%	85%	90%	95%	97.5%	99%	99.5%	99.75%	99.9%	99.95%
Two-sided	50%	60%	70%	80%	90%	95%	98%	99%	99.5%	99.8%	99.9%

Calculating the confidence interval :

Let's say we have a sample with size 11, sample mean 10, and sample variance 2. For 90% confidence with 10 degrees of freedom, the one-sided  value from the table is 1.372 . Then with confidence interval calculated from

\overline{X} _{n} \pm t _{\alpha,\nu} (S _{n})/(\sqrt{n) } ,

we determine that with 90% confidence we have a true mean lying below

10 +1 . 372 (\sqrt{ 2)/(}{ \sqrt{ 11 } }) = 10 . 585 ~.

In other words, 90% of the times that an upper threshold is calculated by this method from particular samples, this upper threshold exceeds the true mean.

And with 90% confidence we have a true mean lying above

10 -1 . 372 (\sqrt{ 2)/(}{ \sqrt{ 11 } }) = 9 . 414 ~.

In other words, 90% of the times that a lower threshold is calculated by this method from particular samples, this lower threshold lies below the true mean.

So that at 80% confidence (calculated from 100% − 2 × (1 − 90%) = 80%), we have a true mean lying within the interval

\left( 10 -1 . 372 (\sqrt{ 2)/(}{ \sqrt{ 11 } }, 10) +1 . 372 (\sqrt{ 2)/(}{ \sqrt{ 11 } } \right)) = ( 9 . 414 , 10 . 585 ) ~.

Saying that 80% of the times that upper and lower thresholds are calculated by this method from a given sample, the true mean is both below the upper threshold and above the lower threshold is not the same as saying that there is an 80% probability that the true mean lies between a particular pair of upper and lower thresholds that have been calculated by this method; see confidence interval and prosecutor's fallacy.

Nowadays, statistical software, such as the R programming language, and functions available in many spreadsheet programs compute values of the  distribution and its inverse without tables.

See also

• F-distribution
• Folded  and half  distributions
• Hotelling's ² distribution
• Multivariate Student distribution
• Standard normal table (Z-distribution table)
•  statistic
• Tau distribution, for internally studentized residuals
• Wilks' lambda distribution
• Wishart distribution
• Modified half-normal distribution^[30] with the pdf on

( 0 , inf ty) is given as f(x) = (2 \beta ^{(\alpha)/(2)} x ^{\alpha-1} \exp( - \beta x ^{2+} \gamma x ))/(\Psi{\left( (\alpha)/(, (\gamma)/(\sqrt{\beta) 2)) \right)}} , where \Psi(\alpha,z) = {} _{1\Psi} ^{1 \left(\begin{matrix}\left(\alpha, (1)/(2)} \right)\\( 1 , 0 )\end{matrix};z \right) denotes the Fox–Wright Psi function.

References

• Senn . S.. Richardson . W.. The first  test. Statistics in Medicine. 13. 8. 785–803. 1994. 8047737 . 10.1002/sim.4780130802.
• Book: Introduction to Mathematical Statistics. Hogg RV, Craig AT. Macmillan. 1978. 4th. New York. B010WFO0SA. Robert V. Hogg.
• Book: Venables . W. N. . B. D. . Ripley . 2002 . Modern Applied Statistics with S . Fourth . Springer .
• Book: Gelman , Andrew . John B. Carlin . Hal S. Stern . Donald B. Rubin . Bayesian Data Analysis . CRC/Chapman & Hall . 2003 . 1-58488-388-X . Second .

External links

• • Earliest Known Uses of Some of the Words of Mathematics (S) (Remarks on the history of the term "Student's distribution")
• First Students on page 112.
• Student's t-Distribution,

Notes and References

1. Web site: Hurst . Simon . The characteristic function of the Student distribution . Financial Mathematics Research Report . No. FMRR006-95 . Statistics Research Report No. SRR044-95 . dead . https://web.archive.org/web/20100218072259/http://wwwmaths.anu.edu.au/research.reports/srr/95/044/ . February 18, 2010 .
2. Norton . Matthew . Khokhlov . Valentyn . Uryasev . Stan . 2019 . Calculating CVaR and bPOE for common probability distributions with application to portfolio optimization and density estimation . . 299 . 1–2 . 1281–1315 . Springer. 10.1007/s10479-019-03373-1 . 1811.11301 . 254231768 . 2023-02-27.
3. Helmert FR . 1875 . Über die Berechnung des wahrscheinlichen Fehlers aus einer endlichen Anzahl wahrer Beobachtungsfehler . de . . 20 . 300–303.
4. Helmert FR . 1876 . Über die Wahrscheinlichkeit der Potenzsummen der Beobachtungsfehler und uber einige damit in Zusammenhang stehende Fragen . de . . 21 . 192–218 .
5. Helmert FR . 1876 . Die Genauigkeit der Formel von Peters zur Berechnung des wahrscheinlichen Beobachtungsfehlers directer Beobachtungen gleicher Genauigkeit . de . The accuracy of Peters' formula for calculating the probable observation error of direct observations of the same accuracy . . 88 . 8–9 . 113–132 . 1876AN.....88..113H . 10.1002/asna.18760880802 .
6. Lüroth J . 1876 . Vergleichung von zwei Werten des wahrscheinlichen Fehlers . de . . 87 . 14 . 209–220 . 1876AN.....87..209L . 10.1002/asna.18760871402 .
7. Pfanzagl J, Sheynin O . 1996 . Studies in the history of probability and statistics. XLIV. A forerunner of the  distribution . . 83 . 4 . 891–898 . 10.1093/biomet/83.4.891 . 1766040 .
8. Sheynin O . 1995 . Helmert's work in the theory of errors . . 49 . 1 . 73–104 . 10.1007/BF00374700 . 121241599.
9. Pearson . K. . 1895 . Contributions to the Mathematical Theory of Evolution. II. Skew Variation in Homogeneous Material . Philosophical Transactions of the Royal Society A: Mathematical, Physical and Engineering Sciences . 186 . 374 . 343–414 . 10.1098/rsta.1895.0010 . free . 1895RSPTA.186..343P . 1364-503X.
10. "Student" pseu. William Sealy Gosset] |year=1908 |title=The probable error of a mean |journal=Biometrika |volume=6 |issue=1 |pages=1–25 |doi=10.1093/biomet/6.1.1 |jstor=2331554 |hdl=10338.dmlcz/143545 |url=http://www.york.ac.uk/depts/maths/histstat/student.pdf.
11. Wendl MC . Michael Christopher Wendl . 2016 . Pseudonymous fame . . 351 . 6280 . 1406 . 10.1126/science.351.6280.1406 . 27013722 . 2016Sci...351.1406W.
12. Book: Mortimer RG . 2005 . Mathematics for Physical Chemistry . 3rd . Elsevier . 9780080492889 . Burlington, MA . 326 . 156200058 . registration.
13. Fisher RA . Ronald Fisher . 1925 . Applications of 'Student's' distribution . Metron . 5 . 90–104 . dead . https://web.archive.org/web/20160305130235/http://www.sothis.ro/user/content/4ef6e90670749a86-student_distribution_1925.pdf . 5 March 2016.
14. Book: Walpole RE, Myers R, Myers S, Ye K . 2006 . Probability & Statistics for Engineers & Scientists . 7th . Pearson . 9788177584042 . New Delhi, IN . 237 . 818811849 .
15. Book: Kruschke, J.K. . John K. Kruschke . 2015 . Doing Bayesian Data Analysis . 2nd . Academic Press . 9780124058880 . 959632184.
16. Book: Casella G, Berger RL . 1990 . Statistical Inference . Duxbury Resource Center . 9780534119584 . 56.
17. Book: Bayesian Analysis for the Social Sciences . limited . Jackman, S. . Wiley Series in Probability and Statistics . Wiley . 2009 . 9780470011546 . 507 . 10.1002/9780470686621.
18. Book: Continuous Univariate Distributions. Johnson NL, Kotz S, Balakrishnan N. Wiley. 1995. 9780471584940. 2nd. 2. Chapter 28.
19. Book: Introduction to Mathematical Statistics. Hogg RV, Craig AT. Macmillan. 1978. 4th. New York. B010WFO0SA. . Sections 4.4 and 4.8. Robert V. Hogg.
20. William Gemmell Cochran. Cochran WG. 1934. The distribution of quadratic forms in a normal system, with applications to the analysis of covariance. Math. Proc. Camb. Philos. Soc.. 30. 2. 178–191. 1934PCPS...30..178C. 10.1017/S0305004100016595. 122547084 .
21. Book: Bayesian Data Analysis . Gelman AB, Carlin JS, Rubin DB, Stern HS . Chapman & Hal l. 1997 . 9780412039911 . 2nd . Boca Raton, FL . 68 .
22. Park SY, Bera AK. 2009. Maximum entropy autoregressive conditional heteroskedasticity model. J. Econom.. 150. 2. 219–230. 10.1016/j.jeconom.2008.12.014.
23. Bailey RW . 1994 . Polar generation of random variates with the  distribution . . 62 . 206 . 779–781 . 10.2307/2153537 . 2153537 . 1994MaCom..62..779B . 120459654 .
24. Book: Families of Frequency Distributions . Ord JK . Griffin . 1972 . 9780852641378 . London, UK . Table 5.1 .
25. Book: Families of frequency distributions . Ord JK . Griffin . 1972 . 9780852641378 . London, UK . Chapter 5.
26. Lange KL, Little RJ, Taylor JM. 1989. Robust Statistical Modeling Using the Distribution. J. Am. Stat. Assoc.. 84. 408. 881–896. 10.1080/01621459.1989.10478852. 2290063.
27. Book: Bayesian Data Analysis. Gelman AB, Carlin JB, Stern HS, Dunson DB, Vehtari A, Rubin DB. CRC Press. 2014. 9781439898208. Boca Raton, Florida. 293. Computationally efficient Markov chain simulation. 3.
28. Shah. Amar . Wilson. Andrew Gordon. Ghahramani. Zoubin. 2014 . Student  processes as alternatives to Gaussian processes. JMLR. 33. Proceedings of the 17th International Conference on Artificial Intelligence and Statistics (AISTATS) 2014, Reykjavik, Iceland. 877–885. 1402.4306 .
29. Chen. Zexun . Wang. Bo. Gorban. Alexander N.. 2019 . Multivariate Gaussian and Student  process regression for multi-output prediction. Neural Computing and Applications. 32 . 8 . 3005–3028 . 10.1007/s00521-019-04687-8. free. 1703.04455 .
30. Sun . Jingchao . Kong . Maiying . Pal . Subhadip . The Modified-Half-Normal distribution: Properties and an efficient sampling scheme . Communications in Statistics - Theory and Methods . 22 June 2021 . 52 . 5 . 1591–1613 . 10.1080/03610926.2021.1934700 . 237919587 . 0361-0926.