Let $ln(X)\sim N(\mu,\sigma^2)$ and $c$ be a constant such that $X+c$ follows a shifted lognormal distribution with parameters $\mu$, $\sigma^2$, and $c$. What I would like to know is the distribution of $\frac{1}{X+c}$. I know that in the special case where $c=0$, $\frac{1}{X}\sim lognormal(-\mu,\sigma^2)$, but what about the more general case?


If $Y = 1/(X+c)$, $c \in \mathbb{R}$, and $X \sim \text{Lognormal}(\mu,\sigma^2)$, then $$ f_Y(y) = f_X(1/y-c)|y^{-2}|. $$ If you're looking for the name of the distribution, I can't help you there. You can see it isn't another lognormal random variable, though.


You can relate this to Johnson's $S_B$-distribution

$$z = \gamma + \delta \log \left(\frac{x-\xi}{\xi+\lambda-x} \right)$$

  • with $\xi = 0$, $\lambda = 1/c$

    from which follows with some rearrancements

    $$\frac{z-\gamma}{\delta}= y = \log \left(\frac{cx}{1-cx} \right) \sim N(-\frac{\gamma}{\delta}, \frac{1}{\delta^2})$$

  • and with $\delta = 1/\sigma$ and $\gamma = \frac{\mu-\ln(c)}{\sigma}$

$$x = \frac{1}{c+e^{-y+\ln(c)}} = \frac{1}{c+e^{y^\prime}}\\ \text{with} \quad y^\prime = \ln(c)-y \sim N\left(\mu=\ln(c)+\frac{\gamma}{\delta},\sigma^2= \frac{1}{\delta^2}\right)$$

You could also see it as a scaled logit-/logistic-normal distribution


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.