2
$\begingroup$

I'm trying to understand how to derive the cumulative distribution function for a lognormal distribution from its probability density function.

I know that the pdf is:

$$f(x) =\frac{e^{-\frac{1}{2}\Bigl(\frac{\ln(x) - \mu}{\sigma}\Bigr)^2}}{x \sigma \sqrt{2\pi}},\ x \gt 0$$

and the cdf is:

$$\Phi(x) = \int_{-\infty}^x f(y) dy = \frac{1}{\sigma\sqrt{2\pi}}\int_0^x e^{-\frac{1}{2}\Bigl(\frac{\ln(y) - \mu}{\sigma}\Bigr)^2}\frac{dy}{y}.$$

Now, I don't know how to get the last formula. I tried the substitution $t=\ln(y)$ but I don't know how to deal with the $1/y$ term.

Am I on the wrong way, or did I make any mistake?

I'd like to know it for my personal knowledge, it's not for a class.

$\endgroup$
  • $\begingroup$ Please add the self-study tag and read its tag wiki, modifying your question as needed to comply with the outline there. $\endgroup$ – Glen_b -Reinstate Monica May 8 '15 at 11:00
6
$\begingroup$

If $Y $ ~ $\log N(\mu,\sigma^2)$ then $$f_Y(y) = \frac{1}{y\sqrt{2\pi \sigma^2}}\exp\left[-\frac{1}{2}\left(\frac{\log y-\mu}{\sigma}\right)^2\right]$$

Now,

$$F_Y(y)=\int_{-\infty}^{y_1}f_Y(y)dy = \int_{-\infty}^{y_1}\frac{1}{y\sqrt{2\pi \sigma^2}}\exp\left[-\frac{1}{2}\left(\frac{\log y-\mu}{\sigma}\right)^2\right]dy$$

$$\frac{\log y - \mu}{\sigma} = z \implies \frac{dy}{y\sigma} = dz, \text{ and } z_1 = \frac{\log y_1 - \mu}{\sigma}$$

$$F_Y(y)=\frac{1}{\sqrt{2\pi}}\int_{-\infty}^{z_1}\exp(-z^2/2)dz$$

Oh boy, that's a standard normal distribution. I think it's easy from here.

$\endgroup$
3
$\begingroup$

I suppose you have the error function in mind.

You can always turn the CDF back into a normal CDF. Recall that $X$ is lognormal iff $\log X\sim N \left(\mu,\sigma^2 \right)$.

$\endgroup$
3
$\begingroup$

1) You need to take account of the Jacobian of the transformation

2) It might be easier to get the pdf from the cdf and then you might spot your error.

$\endgroup$

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.