= Log-normal distribution
{wiki}
= Lognormal distribution
{synonym}
= Lognormal
{synonym}
A positive <Random variable> is <log-normal distribution>[log-normal] if its <logarithm> has a <Normal distribution>. For parameters $\mu\in\mathbb R$ and $\sigma>0$, write $X=e^{\mu+\sigma Z}$ with $Z$ <standard normal distribution>[standard normal]. The <change of variables> $z=(\log x-\mu)/\sigma$ gives the <probability density function>[density]
$$
f_X(x)=\frac{1}{x\sigma\sqrt{2\pi}}
\exp\left[-\frac{(\log x-\mu)^2}{2\sigma^2}\right],\qquad x>0.
$$
The parameters describe the <logarithm>, not the <mean> and <standard deviation> of $X$ itself.
The normal <Moment-generating function> gives every real power <moment>:
$$
\mathbb EX^r=e^{r\mu+r^2\sigma^2/2},\qquad r\in\mathbb R.
$$
Consequently the <mean> is $e^{\mu+\sigma^2/2}$, the <variance> is $(e^{\sigma^2}-1)e^{2\mu+\sigma^2}$, and the <median> is $e^\mu$. For $\mu=0,\sigma=1$, the <median> is one but the <mean> is $e^{1/2}$, illustrating the right-skewed shape.
All power <moments> are finite, yet $\mathbb E e^{tX}=\infty$ for every $t>0$. Indeed the normal-coordinate <integral> contains $\exp(te^{\mu+\sigma z}-z^2/2)$, whose exponential-of-exponential growth defeats the Gaussian tail as $z\to\infty$. Thus <log-normal distribution>[log-normal laws] are heavy-tailed in the positive-exponential-moment convention despite having finite <variance>.
<independent> <log-normal distribution>[log-normal] factors multiply to a <log-normal distribution>[log-normal] variable because their <logarithms> add as <independent> normal variables. This is the natural setting for positive quantities with multiplicative variability. The case $\sigma=0$ is a deterministic positive limit, rather than a law with the displayed <probability density function>[density].
Back to article page