Suppose that $X,Y$ are two independent sub-Gaussian RVs. Let $Z=XY$. Is $Z$ also sub-Gaussian? Can someone provide any reference presenting some basic properties of sub-Gaussian RVs. Thank you in advance!
Asked
Active
Viewed 4,233 times
4
-
http://perso-math.univ-mlv.fr/users/banach/SpringSummerSchool2011/Vershynin-RMT-course-IHP.pdf, http://cnx.org/content/m37185/latest/ and Folger and Rahut's Compressive sensing book all list some basic properties of sub Gaussian RV's. – Batman May 21 '14 at 13:49
1 Answers
7
The product of two sub-gaussian random variables could be as sub-gaussian as not sub-gaussian.
To see that $Z = XY$ is not sub-gaussian in general when $X$ and $Y$ are you just need to take $X, Y \sim \mathcal{N}(0, 1/2)$. Write $$ XY = \frac 1 4 (X + Y)^2 - \frac 1 4 (X - Y)^2 $$ Both $(X+Y)^2$ and $(X-Y)^2$ are distributed according to $\chi^2_1$(see here).
In fact, $\chi^2_p$ distribution is sub-exponential (not sub-gaussian) and linear combination of sub-exponential random variables is also sub-exponential.
As for $Z = XY$ being sub-gaussian take $X \sim \mathcal{N}(0, 1)$ and $Y \sim \mathcal{R}(1/2)$, where
$$ \mathcal{R}(p) = \begin{cases} +1, ~p \\ -1, ~1 - p \end{cases} $$

tortue
- 2,223
-
2+1 In general, it is true that the product of two sub-Gaussian RVs is sub-Exponential. See Lemma 2.7.6 here https://www.math.uci.edu/~rvershyn/papers/HDP-book/HDP-book.pdf – Student Aug 10 '22 at 03:58