The Log-Normal Distribution

A random variable XX is log-normal if ln⁡X\ln X is normally distributed. Formally, if Y=ln⁡X∼N(μ,σ2)Y=\ln X \sim N(\mu,\sigma^2), then X=eYX=e^{Y} is log-normal with parameters μ\mu and σ\sigma, these are the mean and standard deviation of ln⁡X\ln X, not of XX itself. Because XX is the exponential of a normal variable it can only take positive values, and its density is always right-skewed.

Log-normal density for a variable whose log has mean 0 and standard deviation 0.8

Log-normal density for a variable whose log has mean 0 and standard deviation 0.8

Key formulas

Quantity

Formula

Mean

E[X]=eμ+σ2/2E[X]=e^{\mu+\sigma^2/2}

Variance

Var(X)=(eσ2−1)e2μ+σ2\mathrm{Var}(X)=(e^{\sigma^2}-1)e^{2\mu+\sigma^2}

Median

eμe^{\mu}

Mode

eμ−σ2e^{\mu-\sigma^2}


Because σ2>0\sigma^2>0, these three centres always sit in the order mode < median < mean, which is the algebraic statement of the right skew. To find probabilities, convert back to the normal scale by taking logs first:


P(X≤x)=P(ln⁡X≤ln⁡x)=Φ ⁣(ln⁡x−μσ)P(X\le x)=P(\ln X\le \ln x)=\Phi\!\left(\frac{\ln x-\mu}{\sigma}\right)


The step is legitimate because ln⁡\ln is strictly increasing, so the event {X≤x}\{X\le x\} and the event {ln⁡X≤ln⁡x}\{\ln X\le\ln x\} are the same event.

What σ\sigma controls

Holding μ\mu fixed changes nothing about the median, which stays at eμe^{\mu}. Raising σ\sigma moves the mode eμ−σ2e^{\mu-\sigma^2} toward zero and lengthens the right tail. Over the range shown in the figure (0<σ≤10<\sigma\le1), it also lowers and flattens the peak because eσ2/2/(σ2π)e^{\sigma^2/2}/(\sigma\sqrt{2\pi}) decreases on that interval.


The three densities in the figure share one μ\mu and differ only in σ\sigma, so any difference in their shapes has to come from σ\sigma alone.

Probability density functions of three log-normal distributions that share the same \mu and differ only in \sigma, labelled A, B, and C.

Probability density functions of three log-normal distributions that share the same μ\mu and differ only in σ\sigma, labelled A, B, and C.

Worked example

Suppose a component's lifetime XX (in years) satisfies ln⁡X∼N(1.2,0.52)\ln X\sim N(1.2,0.5^2). Find P(X>5)P(X>5).


Step 1, take logs: ln⁡5=1.609\ln 5=1.609. Step 2, standardize: z=1.609−1.20.5=0.82z=\dfrac{1.609-1.2}{0.5}=0.82. Step 3, use the standard normal table: Φ(0.82)≈0.794\Phi(0.82)\approx0.794, so P(X>5)=1−0.794=0.206P(X>5)=1-0.794=0.206. A quick sanity check: the median lifetime is e1.2=3.32e^{1.2}=3.32 years, so 5 years is above the median and the answer must be below 0.5.



The figure shows the probability density functions of three log-normal distributions, drawn as curves A, B and C. All three share μ=0\mu = 0; their σ\sigma values are 0.25, 0.5 and 1.0, in some order. Reading the relative spread of the curves and where each one peaks, which assignment of σ\sigma to the three curves is correct?

Probability density functions of three log-normal distributions that share the same \mu and differ only in \sigma, labelled A, B, and C.

Probability density functions of three log-normal distributions that share the same μ\mu and differ only in σ\sigma, labelled A, B, and C.

Log-normal density for a variable whose log has mean 0 and standard deviation 0.8

Log-normal density for a variable whose log has mean 0 and standard deviation 0.8

Choose an answer

Create a free account to get started

Unlock full course notes by signing up

  • Video lessons taught by top educators
  • Practice questions with guided solutions
  • Written review notes for reference
  • Expert answers anytime from course tutors

If a random variable XX follows a log-normal distribution, which statement below is always true?

Choose an answer

Create a free account to get started

Unlock full course notes by signing up

  • Video lessons taught by top educators
  • Practice questions with guided solutions
  • Written review notes for reference
  • Expert answers anytime from course tutors

The lifetime in years, XX, of a certain sensor is log-normally distributed with ln⁡X∼N(1.5,0.42)\ln X \sim N(1.5, 0.4^2). What is P(X>10)P(X > 10)? Give your answer as a decimal probability.

Create a free account to get started

Unlock full course notes by signing up

  • Video lessons taught by top educators
  • Practice questions with guided solutions
  • Written review notes for reference
  • Expert answers anytime from course tutors