# Studentized range distribution

Jump to: navigation, search
Parameters k > 1 — the number of groups ν > 0 — degrees of freedom q ∈ [0; +∞) ${\displaystyle {\begin{matrix}f(q;k,\nu )={\frac {{\sqrt {2\pi }}k(k-1)\nu ^{\nu /2}}{\Gamma \left({\frac {\nu }{2}}\right)2^{\nu /2-1}}}\int _{0}^{\infty }x^{\nu }\varphi ({\sqrt {\nu }}x)\times \\[0.5em]\left[\int _{-\infty }^{\infty }\varphi (u)\varphi (u-qx)(\Phi (u)-\Phi (u-qx))^{k-2}\,{\text{d}}u\right]\,{\text{d}}x\end{matrix}}}$ ${\displaystyle {\begin{matrix}F(q;k,\nu )={\frac {k\nu ^{\nu /2}}{\Gamma \left({\frac {\nu }{2}}\right)2^{{\frac {\nu }{2}}-1}}}\int _{0}^{\infty }x^{\nu -1}e^{-\nu x^{2}/2}\times \\[0.5em]\left[\int _{-\infty }^{\infty }\varphi (u)(\Phi (u)-\Phi (u-qx))^{k-1}\,{\text{d}}u\right]\,{\text{d}}x\end{matrix}}}$

In probability and statistics, Studentized range distribution is a continuous probability distribution that arises when estimating the range of a normally distributed population in situations where the sample size is small and population standard deviation is unknown.

Suppose that we take a sample of size n from each of k populations with the same normal distribution N(μσ) and suppose that ${\displaystyle {\bar {y}}}$min is the smallest of these sample means and ${\displaystyle {\bar {y}}}$max is the largest of these sample means, and suppose S2 is the pooled sample variance from these samples. Then the following random variable has a Studentized range distribution.

${\displaystyle q={\frac {{\overline {y}}_{\max }-{\overline {y}}_{\min }}{S{\sqrt {2/n}}}}}$

## Definition

### Probability density function

Differentiating the cumulative distribution function with respect to q gives the probability density function.

${\displaystyle f(q;k,\nu )={\frac {{\sqrt {2\pi }}k(k-1)\nu ^{\nu /2}}{\Gamma \left({\frac {\nu }{2}}\right)2^{\nu /2-1}}}\int _{0}^{\infty }x^{\nu }\varphi ({\sqrt {\nu }}x)\left[\int _{-\infty }^{\infty }\varphi (u)\varphi (u-qx)(\Phi (u)-\Phi (u-qx))^{k-2}\,{\text{d}}u\right]\,{\text{d}}x}$

### Cumulative distribution function

The cumulative distribution function is given by [1]

${\displaystyle F(q;k,\nu )={\frac {k\nu ^{\nu /2}}{\Gamma \left(\nu /2\right)2^{\nu /2-1}}}\int _{0}^{\infty }x^{\nu -1}e^{-\nu x^{2}/2}\left[\int _{-\infty }^{\infty }\varphi (u)(\Phi (u)-\Phi (u-qx))^{k-1}\,{\text{d}}u\right]\,{\text{d}}x}$

### Special cases

When the degrees of freedom approach infinity, the standard normal distribution can be used for the general equation above. If k is 2 or 3,[2] the studentized range probability distribution function can be directly evaluated, where ${\displaystyle \varphi (z)}$ is the standard normal probability density function.

${\displaystyle f(q;k=2)={\sqrt {2}}\varphi \left({\frac {q}{\sqrt {2}}}\right)}$
${\displaystyle f(q;k=3)=6{\sqrt {2}}\varphi \left({\frac {q}{\sqrt {2}}}\right)\left(\Phi \left({\frac {q}{\sqrt {6}}}\right)-{\frac {1}{2}}\right)}$

When the degrees of freedom approaches infinity the studentized range cumulative distribution can be calculated at all k using the standard normal distribution.

${\displaystyle F(q;k)=k\int _{-\infty }^{\infty }\varphi (z)[\Phi (z)-\Phi (z-q)]^{k-1}\,{\text{d}}z}$

## How the studentized range distribution arises

For any probability distribution f, the range probability distribution is:[2]

${\displaystyle f(r;k)=k(k-1)\int _{-\infty }^{\infty }f\left(t+{\frac {1}{2}}r\right)f\left(t-{\frac {1}{2}}r\right)\left(\int _{t-{\frac {1}{2}}r}^{t+{\frac {1}{2}}r}f(x)\,{\text{d}}x\right)^{k-2}\,{\text{d}}t}$

What this means, is that we are adding up the likelihood that, given k draws from a distribution, two of them differ by r, and the remaining k-2 draws all fall between the two extreme values. If we use u substitution where ${\displaystyle u=t+{\frac {r}{2}}}$ and define F as the cumulative distribution function of f, then the equation can be simplified.

${\displaystyle f(r;k)=k(k-1)\int _{-\infty }^{\infty }f(u-r)f(u)(F(u)-F(u-r))^{k-2}\,{\text{d}}u}$

In order to create the studentized range distribution, we first use the standard normal distribution for f and F, and change the variable r to q.

${\displaystyle f(q;k)=Sk(k-1)\int _{-\infty }^{\infty }\varphi (u-Sq)\varphi (u)\left[\Phi (u)-\Phi (u-Sq)\right]^{k-2}\,{\text{d}}u}$

The chi distribution is:

${\displaystyle f(x;\nu )={\frac {2^{1-\nu /2}x^{\nu -1}e^{-x^{2}/2}}{\Gamma (\nu /2)}}}$

If we apply a change of variables we see it can also be expressed as:

${\displaystyle f(S;\nu )={\frac {\nu ^{\nu /2}S^{\nu -1}e^{-\nu S^{2}/2}}{2^{\nu /2-1}\Gamma (\nu /2)}}}$

Multiplying the two and integrating over S gives:

${\displaystyle f(q;k,\nu )={\frac {\nu ^{\nu /2}}{2^{\nu /2-1}\Gamma \left({\frac {\nu }{2}}\right)}}\int _{0}^{\infty }S^{\nu }e^{-\nu S^{2}/2}k(k-1)\int _{-\infty }^{\infty }\varphi (u-Sq)\varphi (u)\left[\Phi (u)-\Phi (u-Sq)\right]^{k-2}\,{\text{d}}u\,{\text{d}}S}$

## Uses

Critical values of the studentized range distribution are used in Tukey's range test.

## References

1. ^ Lund, R. E.; Lund, J. R. (1983). "Algorithm AS 190: Probabilities and Upper Quantiles for the Studentized Range". Journal of the Royal Statistical Society. 32 (2): 204–10. JSTOR 2347300.
2. ^ a b A. T. McKay (1933). "A Note on the Distribution of Range in Samples of n". Biometrika. 25 (3): 415–20. JSTOR 2332292.