Talk:Gaussian measure

What about this

What about measures of the form

$\exp \left({\frac {-1}{2}}x^{T}A^{-1}x\right)\,dx$

where x ∈ R ⁿ, A is a positive-definite matrix that is not necessarily a scalar multiple of the identity matrix, and dx means Lebesgue measure? Are those not considered "Gaussian measures"? Michael Hardy (talk) 23:41, 19 October 2008 (UTC)

Yes, they are (but you have to normalize it). Why do you ask? An image of a Gaussian measure under a linear (or affine) transformation is a Gaussian measure. Boris Tsirelson (talk) 12:29, 20 October 2008 (UTC)

Merging with normal distribution?

Is there a case for merging this page with the page on normal distributions? These are just two different names for the same object... Hairer (talk) 23:11, 14 November 2012 (UTC)

Infinite-dimensional Gaussian measures are seldom called normal distributions. Boris Tsirelson (talk) 06:00, 15 November 2012 (UTC)

You are right, good point. Hairer (talk) 08:47, 15 November 2012 (UTC)

Order of a random variable

Hi Boris. Noticed that you redid my edit. I am not an experienced mathematician, but know some. I have not heard of the order of a random variable before, which was the reason for my edit. Maybe an article on what that is is needed or should be linked to to clarify? Ravn (talk) 06:04, 21 October 2014 (UTC)

Hi. Really, you are right if we treat that claim (in the lead) as a mathematical theorem. If you insists on that, do it; I'll not revert more. But it is written "loosely speaking", thus I treat it as rather a hint, some intuition toward the theorem. (I did not write that phrase; someone else did.) Surely it is meant that typical values of that random variable are of order sqrt(N).

But if you insist on your formulation in terms of variance, please note that the variance of the sum is the sum of variances, thus, of order N, not sqrt(N). Do not confuse variance with mean square deviation. Boris Tsirelson (talk) 11:37, 21 October 2014 (UTC)