Definition of Log-normal distribution
The continuous random variable X has a log-normal distribution if
the random variable Y=ln(X) has a normal distribution with mean
μ and standard deviation σ. The probability density
function of X is
f(x)=1√2πσxe−12σ2(lnx−μ)2,x≥0
In Log-normal distribution μ is called location parameter, since it locates the curve of the distribution, and σ is called scale parameter, since the shape of the curve depends on the value of σ.
Notation : X∼LN(μ,σ2).
Standard Log-Normal Distribution
The standard form of log-normal distribution is obtained by taking μ=0 and σ=1. The p.d.f of standard log-normal distribution is
f(x)=1√2πxe−12(lnx)2;x≥0
Moments of Log-normal distribution
The rth raw moment of log-normal distribution is
μ′r=eμr+12r2σ2.
Proof
Let Y=logX∼N(μ,σ2). So X=eY∼LN(μ,σ2). Hence, the rth raw moment of log-normal distribution is
μ′r=E(Xr)=E(erY)=MY(r)(∵
Hence, for r=1,
\begin{aligned} \mu_1^\prime &= e^{\mu+\frac{1}{2}\sigma^2} \end{aligned}
and
for r=2, \mu_2^\prime = e^{2\mu+2\sigma^2}.
\begin{aligned} \text{ Variance = } \mu_2 &= \mu_2^\prime-(\mu_1^\prime)^2\\ & = e^{2\mu+2\sigma^2}-e^{2\mu+\sigma^2}\\ & = e^{2\mu+\sigma^2}(e^{\sigma^2}-1). \end{aligned}
Quartiles of Log-normal distribution
The quartiles of log-normal distribution are
Q_1= e^{\mu -0.675\sigma }
,
Q_2 = e^{\mu -0\sigma }=e^\mu
,
Q_3 = e^{\mu +0.675\sigma }
.
Proof
Let X has log-normal distribution with parameter \mu and \sigma. Then Y=\log_e X \sim N(\mu, \sigma^2) distribution. Hence Z=\frac{\log_e X -\mu}{\sigma}\sim N(0,1) distribution.
The i^{th} quartile Q_i is given by
\begin{aligned} & P(X\leq Q_i) = \frac{i}{4}\\ \Rightarrow & P\bigg(\frac{\log_e X-\mu}{\sigma}\leq\frac{\log_e Q_i-\mu}{\sigma}\bigg) = \frac{i}{4}\\ \Rightarrow & P\bigg(Z\leq\frac{\log_e Q_i-\mu}{\sigma}\bigg) = \frac{i}{4}. \end{aligned}
For i=1, the first quartile Q_1 is given by
\begin{aligned} & P\bigg(Z\leq\frac{\log_e Q_1-\mu}{\sigma}\bigg) = \frac{1}{4}=0.25\\ \Rightarrow & \frac{\log_e Q_1-\mu}{\sigma}= z_{0.25}\\ \Rightarrow & Q_1 = e^{\mu -0.675\sigma }. \end{aligned}
For i=2, the second quartile Q_2=\text{ median} is given by
\begin{aligned} & P\bigg(Z\leq\frac{\log_e Q_2-\mu}{\sigma}\bigg) = \frac{2}{4}=0.5\\ \Rightarrow & \frac{\log_e Q_2-\mu}{\sigma}= z_{0.5}\\ \Rightarrow & Q_2 = e^{\mu -0\sigma }=e^\mu. \end{aligned}
For i=3, the third quartile Q_3 is given by
\begin{aligned} & P\bigg(Z\leq\frac{\log_e Q_3-\mu}{\sigma}\bigg) = \frac{3}{4}=0.75\\ \Rightarrow & \frac{\log_e Q_3-\mu}{\sigma}= z_{0.75}\\ \Rightarrow & Q_3 = e^{\mu +0.675\sigma }. \end{aligned}
Mode of Log-normal distribution
The mode of log-normal distribution is e^{\mu - \sigma^2}.
Proof
Mode of log-normal distribution can be obtained by solving f^\prime(x)=0 and f^{\prime\prime}<0.
The p.d.f. of log-normal distribution is
\begin{equation*} f(x;\mu,\sigma) = \frac{1}{\sqrt{2\pi}\sigma x}e^{-\frac{1}{2\sigma^2}(\ln x -\mu)^2},\; x\geq 0 \\ \end{equation*}
Differentiating above density function withe respect to x and
equating to zero, we get
\begin{aligned} & f^\prime(x)=0\\ &\Rightarrow \frac{1}{\sqrt{2\pi}\sigma x}e^{-\frac{1}{2\sigma^2}(\ln x -\mu)^2}\times\big( -\frac{1}{2\sigma^2}2(\ln x -\mu)\big) \times \frac{1}{x}\\ & + e^{-\frac{1}{2\sigma^2}(\ln x -\mu)^2}\times\frac{1}{\sqrt{2\pi}\sigma}\big(-\frac{1}{x^2}\big) =0\\ &\Rightarrow \frac{-f(x)}{x} \bigg[\frac{\ln x - \mu +\sigma^2}{\sigma^2}\bigg]=0\\ &\Rightarrow \ln x = \mu - \sigma^2\\ &\Rightarrow x = e^{\mu - \sigma^2}. \end{aligned}
Also, f^{\prime\prime}<0. Hence the mode of the log-normal distribution is e^{\mu -\sigma^2}.