TransWikia.com

What happens to the Gaussian as $sigma to infty$ at $x to infty$?

Cross Validated Asked by rkabra on February 6, 2021

Does the Gaussian $N(0, sigma^2)$ have a limiting distribution as $sigma to infty$? And does the density in fact decay to 0 as $x to infty$?

If we want the ratio of the density at the mean over the density as $x$ also approaches infinity, we can set up the following iterated limit:

$$L := lim_{sigma to infty, x to infty} f(x)/f(0) = exp(-x^2/2sigma^2)$$
$$log L = lim_{sigma to infty, x to infty} -x^2/2sigma^2$$

But can we solve for L or is it indeterminate? A cursory reading of Ivlev and Shilin (2014) suggests it is in fact indeterminate.

Reference:
Ivlev, V. V., and I. A. Shilin. "On a generalization of l’Hopital’s rule for multivariable functions." arXiv preprint arXiv:1403.3006 (2014).

One Answer

The limit $log L$ is indeterminate because it will depend on the path that $(sigma,x)$ goes to $(infty,infty)$.

  • For instance if $x = n^2$ and $sigma = n$ then both go to infinity for $n to infty$ and the limit $log L$ will be infinity.
  • But you can also make the limit $log L$ go to zero when you choose instead $x = n$ and $sigma = n^2$.

Note that the limit of the normal distribution for $sigma to infty$ is not a probability density function (because the limit is zero everywhere). What do you wanna do with it? What is it supposed to mean?

Answered by Sextus Empiricus on February 6, 2021

Add your own answers!

Ask a Question

Get help from others!

© 2024 TransWikia.com. All rights reserved. Sites we Love: PCI Database, UKBizDB, Menu Kuliner, Sharing RPP