Log-normal distribution

From Wikipedia, the free encyclopedia - View original article

Log-normal
Probability density function
Plot of the Lognormal PDF
Some log-normal density functions with identical location parameter μ but differing scale parameters σ
Cumulative distribution function
Plot of the Lognormal CDF
Cumulative distribution function of the log-normal distribution (with μ = 0 )
Notation\ln\mathcal{N}(\mu,\,\sigma^2)
Parametersσ2 > 0 — shape (real),
μR — log-scale
Supportx ∈ (0, +∞)
pdf\frac{1}{x\sqrt{2\pi}\sigma}\ e^{-\frac{\left(\ln x-\mu\right)^2}{2\sigma^2}}
CDF\frac12 + \frac12\,\mathrm{erf}\Big[\frac{\ln x-\mu}{\sqrt{2}\sigma}\Big]
Meane^{\mu+\sigma^2/2}
Mediane^{\mu}\,
Modee^{\mu-\sigma^2}
Variance(e^{\sigma^2}\!\!-1) e^{2\mu+\sigma^2}
Skewness(e^{\sigma^2}\!\!+2) \sqrt{e^{\sigma^2}\!\!-1}
Ex. kurtosise^{4\sigma^2}\!\! + 2e^{3\sigma^2}\!\! + 3e^{2\sigma^2}\!\! - 6
Entropy\frac12 + \frac12 \ln(2\pi\sigma^2) + \mu
MGF(defined only on the negative half-axis, see text)
CFrepresentation \sum_{n=0}^{\infty}\frac{(it)^n}{n!}e^{n\mu+n^2\sigma^2/2} is asymptotically divergent but sufficient for numerical purposes
Fisher information\begin{pmatrix}1/\sigma^2&0\\0&1/(2\sigma^4)\end{pmatrix}
 
Jump to: navigation, search
Log-normal
Probability density function
Plot of the Lognormal PDF
Some log-normal density functions with identical location parameter μ but differing scale parameters σ
Cumulative distribution function
Plot of the Lognormal CDF
Cumulative distribution function of the log-normal distribution (with μ = 0 )
Notation\ln\mathcal{N}(\mu,\,\sigma^2)
Parametersσ2 > 0 — shape (real),
μR — log-scale
Supportx ∈ (0, +∞)
pdf\frac{1}{x\sqrt{2\pi}\sigma}\ e^{-\frac{\left(\ln x-\mu\right)^2}{2\sigma^2}}
CDF\frac12 + \frac12\,\mathrm{erf}\Big[\frac{\ln x-\mu}{\sqrt{2}\sigma}\Big]
Meane^{\mu+\sigma^2/2}
Mediane^{\mu}\,
Modee^{\mu-\sigma^2}
Variance(e^{\sigma^2}\!\!-1) e^{2\mu+\sigma^2}
Skewness(e^{\sigma^2}\!\!+2) \sqrt{e^{\sigma^2}\!\!-1}
Ex. kurtosise^{4\sigma^2}\!\! + 2e^{3\sigma^2}\!\! + 3e^{2\sigma^2}\!\! - 6
Entropy\frac12 + \frac12 \ln(2\pi\sigma^2) + \mu
MGF(defined only on the negative half-axis, see text)
CFrepresentation \sum_{n=0}^{\infty}\frac{(it)^n}{n!}e^{n\mu+n^2\sigma^2/2} is asymptotically divergent but sufficient for numerical purposes
Fisher information\begin{pmatrix}1/\sigma^2&0\\0&1/(2\sigma^4)\end{pmatrix}

In probability theory, a log-normal distribution is a continuous probability distribution of a random variable whose logarithm is normally distributed. Thus, if the random variable X is log-normally distributed, then Y = \log(X) has a normal distribution. Likewise, if Y has a normal distribution, then X = \exp(Y) has a log-normal distribution. A random variable which is log-normally distributed takes only positive real values.

Log-normal is also written log normal or lognormal. The distribution is occasionally referred to as the Galton distribution or Galton's distribution, after Francis Galton.[1] The log-normal distribution also has been associated with other names, such as McAlister, Gibrat and Cobb–Douglas.[1]

A variable might be modeled as log-normal if it can be thought of as the multiplicative product of many independent random variables each of which is positive. (This is justified by considering the central limit theorem in the log-domain.) For example, in finance, the variable could represent the compound return from a sequence of many trades (each expressed as its return + 1); or a long-term discount factor can be derived from the product of short-term discount factors. In wireless communication, the sas caused by shadowing or slow fading from random objects is often assumed to be log-normally distributed: see log-distance path loss model.

The log-normal distribution is the maximum entropy probability distribution for a random variate X for which the mean and variance of \ln(X) are fixed.[2]

μ and σ[edit]

In a log-normal distribution X, the parameters denoted μ and σ are, respectively, the mean and standard deviation of the variable’s natural logarithm (by definition, the variable’s logarithm is normally distributed), which means

 X=e^{\mu+\sigma Z}

with Z a standard normal variable.

This relationship is true regardless of the base of the logarithmic or exponential function. If loga(Y) is normally distributed, then so is logb(Y), for any two positive numbers ab ≠ 1. Likewise, if e^X is log-normally distributed, then so is a^{X}, where a is a positive number ≠ 1.

On a logarithmic scale, μ and σ can be called the location parameter and the scale parameter, respectively.

In contrast, the mean and standard deviation of the non-logarithmized sample values are denoted m and s.d. in this article.

A log-normal distribution with mean m and variance v has parameters[3]

 \mu=\ln\left(\frac{m^2}{\sqrt{v+m^2}}\right), \sigma=\sqrt{\ln\left(1+\frac{v}{m^2}\right)}

Characterization[edit]

Probability density function[edit]

The probability density function of a log-normal distribution is:[1]

f_X(x;\mu,\sigma) = \frac{1}{x \sigma \sqrt{2 \pi}}\, e^{-\frac{(\ln x - \mu)^2}{2\sigma^2}},\ \ x>0

This follows by applying the change-of-variables rule on the density function of a normal distribution.

Cumulative distribution function[edit]

The cumulative distribution function is

F_X(x;\mu,\sigma) = \frac12 \left[ 1 + \operatorname{erf}\!\left(\frac{\ln x - \mu}{\sigma\sqrt{2}}\right) \right] = \frac12 \operatorname{erfc}\!\left(-\frac{\ln x - \mu}{\sigma\sqrt{2}}\right) = \Phi\bigg(\frac{\ln x - \mu}{\sigma}\bigg),

where erfc is the complementary error function, and Φ is the cumulative distribution function of the standard normal distribution.

Characteristic function and moment generating function[edit]

All moments of the log-normal distributions exist and it holds that

\operatorname{E}(X^n)=\mathrm{e}^{n\mu+\frac{n^2\sigma^2}{2}}.

However, the moment generating function

\operatorname{E}(e^{t X})=\sum_{n=0}^\infty \frac{t^n}{n!}\operatorname{E}(X^n)

does not converge.

The characteristic function, E[e itX], has a number of representations.[citation needed] The integral itself converges for Im(t) ≤ 0. The simplest representation is obtained by Taylor expanding e itX and using formula for moments below, giving[citation needed]

\varphi(t) = \sum_{n=0}^\infty \frac{(it)^n}{n!}e^{n\mu+n^2\sigma^2/2}.

This series representation is divergent for Re(σ2) > 0.[citation needed] However, it is sufficient for evaluating the characteristic function numerically at positive \sigma as long as the upper limit in the sum above is kept bounded, n ≤ N, where

\max(|t|,|\mu|) \ll N \ll \frac{2}{\sigma^2}\ln\frac{2}{\sigma^2}

and σ2 < 0.1.[citation needed] To bring the numerical values of parameters μσ into the domain where strong inequality holds true one could use the fact that if X is log-normally distributed then Xm is also log-normally distributed with parameters μmσm. Since  \mu\sigma^2 \propto m^3, the inequality could be satisfied for sufficiently small m. The sum of series first converges to the value of φ(t) with arbitrary high accuracy if m is small enough, and left part of the strong inequality is satisfied. If considerably larger number of terms are taken into account the sum eventually diverges when the right part of the strong inequality is no longer valid.

Another useful representation is available[4][5] by means of double Taylor expansion of e(ln x − μ)2/(2σ2).

The moment-generating function for the log-normal distribution does not exist on the domain R, but only exists on the half-interval (−∞, 0].[citation needed]

Properties[edit]

Location and scale[edit]

For the log-normal distribution, the location and scale properties of the distribution are more readily treated using the geometric mean and geometric standard deviation than the arithmetic mean and standard deviation.

Geometric moments[edit]

The geometric mean of the log-normal distribution is e^{\mu}. Because the log of a log-normal variable is symmetric and quantiles are preserved under monotonic transformations, the geometric mean of a log-normal distribution is equal to its median.[6]

The geometric mean (mg) can alternatively be derived from the arithmetic mean (ma) in a log-normal distribution by:

 m_g = m_ae^{-\tfrac{1}{2}\sigma^2}.

Note that the geometric mean is less than the arithmetic mean. This is due to the AM–GM inequality, and corresponds to the logarithm being convex down. The correction term e^{-\tfrac{1}{2}\sigma^2} can accordingly be interpreted as a convexity correction. From the point of view of stochastic calculus, this is the same correction term as in Itō's lemma for geometric Brownian motion.

The geometric standard deviation is equal to e^{\sigma}.[citation needed]

Arithmetic moments[edit]

If X is a lognormally distributed variable, its expected value (E – the arithmetic mean), variance (Var), and standard deviation (s.d.) are

\begin{align}   & \operatorname{E}[X] = e^{\mu + \tfrac{1}{2}\sigma^2}, \\   & \operatorname{Var}[X] = (e^{\sigma^2} - 1) e^{2\mu + \sigma^2} = (e^{\sigma^2} - 1)(\operatorname{E}[X])^2\\   & \operatorname{s.d.}[X] = \sqrt{\operatorname{Var}[X]} = e^{\mu + \tfrac{1}{2}\sigma^2}\sqrt{e^{\sigma^2} - 1}.   \end{align}

Equivalently, parameters μ and σ can be obtained if the expected value and variance are known; it is simpler if σ is computed first:

\begin{align}   \mu &= \ln(\operatorname{E}[X]) - \frac12 \ln\!\left(1 + \frac{\mathrm{Var}[X]}{(\operatorname{E}[X])^2}\right) = \ln(\operatorname{E}[X]) - \frac12 \sigma^2, \\   \sigma^2 &= \ln\!\left(1 + \frac{\operatorname{Var}[X]}{(\operatorname{E}[X])^2}\right).   \end{align}

For any real or complex number s, the sth moment of log-normal X is given by[1]

\operatorname{E}[X^s] = e^{s\mu + \tfrac{1}{2}s^2\sigma^2}.

A log-normal distribution is not uniquely determined by its moments E[Xk] for k ≥ 1, that is, there exists some other distribution with the same moments for all k.[1] In fact, there is a whole family of distributions with the same moments as the log-normal distribution.[citation needed]

Mode and median[edit]

Comparison of mean, median and mode of two log-normal distributions with different skewness.

The mode is the point of global maximum of the probability density function. In particular, it solves the equation (ln ƒ)′ = 0:

\mathrm{Mode}[X] = e^{\mu - \sigma^2}.

The median is such a point where FX = 1/2:

\mathrm{Med}[X] = e^\mu\,.

Coefficient of variation[edit]

The coefficient of variation is the ratio s.d. over m (on the natural scale) and is equal to:

\sqrt{e^{\sigma^2}\!\!-1}

Partial expectation[edit]

The partial expectation of a random variable X with respect to a threshold k is defined as  g(k) = \int_k^\infty \!xf(x)\, dx where  f(x) is the probability density function of X. Alternatively, and using the definition of conditional expectation, it can be written as g(k)=E[X | X > k]*P(X > k). For a log-normal random variable the partial expectation is given by:

g(k) = \int_k^\infty \!xf(x)\, dx             = e^{\mu+\tfrac{1}{2}\sigma^2}\, \Phi\!\left(\frac{\mu+\sigma^2-\ln k}{\sigma}\right).

The derivation of the formula is provided in the discussion of this Wikipedia entry. The partial expectation formula has applications in insurance and economics, it is used in solving the partial differential equation leading to the Black–Scholes formula.

Other[edit]

A set of data that arises from the log-normal distribution has a symmetric Lorenz curve (see also Lorenz asymmetry coefficient).[7]

The harmonic (H), geometric (G) and arithmetic (A) means of this distribution are related;[8] such relation is given by

H = \frac{G^2}{ A} .

Log-normal distributions are infinitely divisible.[1]

Occurrence[edit]

The log-normal distribution is important in the description of natural phenomena. The reason is that for many natural processes of growth, growth rate is independent of size. This is also known as Gibrat's law, after Robert Gibrat (1904–1980) who formulated it for firms (companies). It can be shown that a growth process following Gibrat's law will result in entity sizes with a log-normal distribution.[9] Examples include:

Fitted cumulative log-normal distribution to annually maximum 1-day rainfalls, see distribution fitting
Consequently, reference ranges for measurements in healthy individuals are more accurately estimated by assuming a log-normal distribution than by assuming a symmetric distribution about the mean.

Maximum likelihood estimation of parameters[edit]

For determining the maximum likelihood estimators of the log-normal distribution parameters μ and σ, we can use the same procedure as for the normal distribution. To avoid repetition, we observe that

f_L (x;\mu, \sigma) = \prod_{i=1}^n \left(\frac 1 x_i\right) \, f_N (\ln x; \mu, \sigma)

where by ƒL we denote the probability density function of the log-normal distribution and by ƒN that of the normal distribution. Therefore, using the same indices to denote distributions, we can write the log-likelihood function thus:

 \begin{align} \ell_L (\mu,\sigma | x_1, x_2, \dots, x_n)   & {} = - \sum _k \ln x_k + \ell_N (\mu, \sigma | \ln x_1, \ln x_2, \dots, \ln x_n) \\ & {} = \operatorname {constant} + \ell_N (\mu, \sigma | \ln x_1, \ln x_2, \dots, \ln x_n). \end{align}

Since the first term is constant with regard to μ and σ, both logarithmic likelihood functions, L and N, reach their maximum with the same μ and σ. Hence, using the formulas for the normal distribution maximum likelihood parameter estimators and the equality above, we deduce that for the log-normal distribution it holds that

\widehat \mu = \frac {\sum_k \ln x_k} n,         \widehat \sigma^2 = \frac {\sum_k \left( \ln x_k - \widehat \mu \right)^2} {n}.

Multivariate log-normal[edit]

If \boldsymbol X \sim \mathcal{N}(\boldsymbol\mu,\,\boldsymbol\Sigma) is a multivariate normal distribution then \boldsymbol Y=\exp(\boldsymbol X) has a multivariate log-normal distribution[22] with mean

\operatorname{E}[\boldsymbol Y]_i=e^{\mu_i+\frac{1}{2}\Sigma_{ii}} ,

and covariance matrix

\operatorname{Var}[\boldsymbol Y]_{ij}=e^{\mu_i+\mu_j + \frac{1}{2}(\Sigma_{ii}+\Sigma_{jj}) }( e^{\Sigma_{ij}} - 1) .

Generating log-normally distributed random variates[edit]

Given a random variate Z drawn from the normal distribution with 0 mean and 1 standard deviation, then the variate

X= e^{\mu + \sigma Z}\,

has a log-normal distribution with parameters \mu and \sigma.

Related distributions[edit]

Y \sim \operatorname{Log-\mathcal{N}}\Big(\textstyle \sum_{j=1}^n\mu_j,\ \sum_{j=1}^n \sigma_j^2 \Big).
\begin{align}   \sigma^2_Z &= \log\!\left[ \frac{\sum e^{2\mu_j+\sigma_j^2}(e^{\sigma_j^2}-1)}{(\sum e^{\mu_j+\sigma_j^2/2})^2} + 1\right], \\   \mu_Z &= \log\!\left[ \sum e^{\mu_j+\sigma_j^2/2} \right] - \frac{\sigma^2_Z}{2}.   \end{align}

In the case that all X_j have the same variance parameter \sigma_j=\sigma, these formulas simplify to

\begin{align}   \sigma^2_Z &= \log\!\left[ (e^{\sigma^2}-1)\frac{\sum e^{2\mu_j}}{(\sum e^{\mu_j})^2} + 1\right], \\   \mu_Z &= \log\!\left[ \sum e^{\mu_j} \right] + \frac{\sigma^2}{2} -  \frac{\sigma^2_Z}{2}.   \end{align}

Similar distributions[edit]

A substitute for the log-normal whose integral can be expressed in terms of more elementary functions[24] can be obtained based on the logistic distribution to get an approximation for the CDF

 F(x;\mu,\sigma) = \left[\left(\frac{e^\mu}{x}\right)^{\pi/(\sigma \sqrt{3})} +1\right]^{-1}.

This is a log-logistic distribution.

See also[edit]

Notes[edit]

  1. ^ a b c d e f Johnson, Norman L.; Kotz, Samuel; Balakrishnan, N. (1994), "14: Lognormal Distributions", Continuous univariate distributions. Vol. 1, Wiley Series in Probability and Mathematical Statistics: Applied Probability and Statistics (2nd ed.), New York: John Wiley & Sons, ISBN 978-0-471-58495-7, MR 1299979 
  2. ^ Park, Sung Y.; Bera, Anil K. (2009). "Maximum entropy autoregressive conditional heteroskedasticity model". Journal of Econometrics (Elsevier): 219–230. Retrieved 2011-06-02. 
  3. ^ "Lognormal mean and variance"
  4. ^ Leipnik, Roy B. (1991), "On Lognormal Random Variables: I – The Characteristic Function", Journal of the Australian Mathematical Society Series B, 32, 327–347.
  5. ^ a b Daniel Dufresne (2009), SUMS OF LOGNORMALS, Centre for Actuarial Studies, University of Melbourne.
  6. ^ Leslie E. Daly, Geoffrey Joseph Bourke (2000) Interpretation and uses of medical statistics Edition: 5. Wiley-Blackwell ISBN 0-632-04763-1, ISBN 978-0-632-04763-5 (page 89)
  7. ^ Damgaard, Christian; Weiner, Jacob (2000). "Describing inequality in plant size or fecundity". Ecology 81 (4): 1139–1142. doi:10.1890/0012-9658(2000)081[1139:DIIPSO]2.0.CO;2. 
  8. ^ Rossman LA (1990) "Design stream flows based on harmonic means". J Hydraulic Engineering ASCE 116 (7) 946–950
  9. ^ Sutton, J. (1997), "Gibrat's Legacy", Journal of Economic Literature XXXV, 40–59.
  10. ^ Huxley, Julian S. (1932). Problems of relative growth. London. ISBN 0-486-61114-0. OCLC 476909537. 
  11. ^ a b WB, Wang; CF Wang, ZN Wu and RF Hu (2013). "Modelling the spreading rate of controlled communicable epidemics through an entropy-based thermodynamic model" 56 (11). SCIENCE CHINA Physics, Mechanics & Astronomy. pp. 2143–2150. 
  12. ^ Makuch, Robert W.; D.H. Freeman, M.F. Johnson (1979). "Justification for the lognormal distribution as a model for blood pressure". Journal of Chronic Diseases 32 (3): 245–250. doi:10.1016/0021-9681(79)90070-5. Retrieved 27 February 2012. 
  13. ^ Ritzema (ed.), H.P. (1994). Frequency and Regression Analysis. Chapter 6 in: Drainage Principles and Applications, Publication 16, International Institute for Land Reclamation and Improvement (ILRI), Wageningen, The Netherlands. pp. 175–224. ISBN 90-70754-33-9. 
  14. ^ Clementi, Fabio; Gallegati, Mauro (2005) "Pareto's law of income distribution: Evidence for Germany, the United Kingdom, and the United States", EconWPA
  15. ^ Black, Fischer and Myron Scholes, "The Pricing of Options and Corporate Liabilities", Journal of Political Economy, Vol. 81, No. 3, (May/June 1973), pp. 637–654.
  16. ^ Madelbrot, Beniot (2004). The (mis-)Behaviour of Markets. 
  17. ^ Bunchen, P., Advanced Option Pricing, University of Sydney coursebook, 2007
  18. ^ O'Connor, Patrick; Kleyner, Andre (2011). Practical Reliability Engineering. John Wiley & Sons. p. 35. ISBN 978-0-470-97982-2. 
  19. ^ http://wireless.per.nl/reference/chaptr03/shadow/shadow.htm[dead link]
  20. ^ Steele, C. (2008). "Use of the lognormal distribution for the coefficients of friction and wear". Reliability Engineering & System Safety 93 (10): 1574–2013. doi:10.1016/j.ress.2007.09.005.  edit
  21. ^ a b Wu, Z.N. (2003), [ "Prediction of the size distribution of secondary ejected droplets by crown splashing of droplets impinging on a solid wall"]. Probabilistic Engineering Mechanics, Volume 18, Issue 3, July 2003, Pages 241–249. doi:10.1016/S0266-8920(03)00028-6
  22. ^ Tarmast, Ghasem (2001) "Multivariate Log–Normal Distribution" ISI Proceedings: Seoul 53rd Session 2001
  23. ^ Gao, X.; Xu, H; Ye, D. (2009), "Asymptotic Behaviors of Tail Density for Sum of Correlated Lognormal Variables". International Journal of Mathematics and Mathematical Sciences, vol. 2009, Article ID 630857. doi:10.1155/2009/630857
  24. ^ Swamee, P. K. (2002). "Near Lognormal Distribution". Journal of Hydrologic Engineering 7 (6): 441–444. doi:10.1061/(ASCE)1084-0699(2002)7:6(441).  edit

References[edit]

Further reading[edit]

External links[edit]