I found here that

The cdf for the max is the cdf for the normal raised to the power of the sample size. [say, $n$]

Since the cdf for the normal is the error function,

$$F_X(x)=\frac{1}{\sigma\sqrt{ 2\pi}} \int_{-\infty}^x \exp{\left(\frac{-(t -\mu)^2}{2\sigma^2} \right)}\mathrm dt=\frac{1}{2}\Bigg[1+ \mathrm{erf} \left( \frac{x-\mu}{\sigma \sqrt 2}\right)\Bigg],$$

and the pdf of the standard normal is

$$f(x) = \frac{e^{-x^{2}/2}} {\sqrt{2\pi}}$$

I presume there is no other workaround to

$$f_{\max(\mathrm X)}(x_i)=n \quad \frac{e^{-x^{2}/2}} {\sqrt{2\pi}}\quad \frac{1}{2}\Bigg[1+ \mathrm{erf} \left( \frac{x-\mu}{\sigma \sqrt 2}\right)\Bigg]^{n-1}$$

Would this be correct? It doesn't get any simpler?

  • $\begingroup$ @GordonSmyth Corrected now. $\endgroup$ Commented Feb 4, 2018 at 21:49

1 Answer 1


Yes it is. Some related ideas/references/bounds are disused in



Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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