Skip to main content
more enticing title and spacing
Source Link
user44143
user44143

Upper bounds on probability concerning maxima of Given correlated Gaussian random variables, how to bound the probability that the first is the largest?

Let $Z\sim \mathcal{N}(\mu, \Sigma)$ be a Gaussian random vector in $\mathbb{R}^d$.

What are some nontrivial bounds on $p=\mathbb{P}(S)$, where $S$ is the event $Z_1=\max_i Z_i$?

This is motivated by the inequality

$$\mathbb{E}(|Z_1|\mathbf{1_S})\ge \max\bigg(0,p\mathbb{E}|Z_1|-\sqrt{\mu_i^2+\Sigma_{ii}-(\mathbb{E}|Z_1|)^2}\sqrt{p-p^2}\bigg)$$$$\mathbb{E}(|Z_1|\mathbf{1_S})\ge \max\bigg(0,p\,\mathbb{E}|Z_1|-\sqrt{\mu_i^2+\Sigma_{ii}-(\mathbb{E}|Z_1|)^2}\sqrt{p-p^2}\bigg)$$

Upper bounds on probability concerning maxima of correlated Gaussian random variables

Let $Z\sim \mathcal{N}(\mu, \Sigma)$ be a Gaussian random vector in $\mathbb{R}^d$.

What are some nontrivial bounds on $p=\mathbb{P}(S)$, where $S$ is the event $Z_1=\max_i Z_i$?

This is motivated by the inequality

$$\mathbb{E}(|Z_1|\mathbf{1_S})\ge \max\bigg(0,p\mathbb{E}|Z_1|-\sqrt{\mu_i^2+\Sigma_{ii}-(\mathbb{E}|Z_1|)^2}\sqrt{p-p^2}\bigg)$$

Given correlated Gaussian random variables, how to bound the probability that the first is the largest?

Let $Z\sim \mathcal{N}(\mu, \Sigma)$ be a Gaussian random vector in $\mathbb{R}^d$.

What are some nontrivial bounds on $p=\mathbb{P}(S)$, where $S$ is the event $Z_1=\max_i Z_i$?

This is motivated by the inequality

$$\mathbb{E}(|Z_1|\mathbf{1_S})\ge \max\bigg(0,p\,\mathbb{E}|Z_1|-\sqrt{\mu_i^2+\Sigma_{ii}-(\mathbb{E}|Z_1|)^2}\sqrt{p-p^2}\bigg)$$

added 4 characters in body
Source Link

Let $Z\sim \mathcal{N}(\mu, \Sigma)$ be a Gaussian random vector in $\mathbb{R}^d$.

What are some nontrivial bounds on $p=\mathbb{P}(S)$, where $S$ is the event $Z_1=\max Z$$Z_1=\max_i Z_i$?

This is motivated by the inequality

$$\mathbb{E}(|Z_1|\mathbf{1_S})\ge \max\bigg(0,p\mathbb{E}|Z_1|-\sqrt{\mu_i^2+\Sigma_{ii}-(\mathbb{E}|Z_1|)^2}\sqrt{p-p^2}\bigg)$$

Let $Z\sim \mathcal{N}(\mu, \Sigma)$ be a Gaussian random vector in $\mathbb{R}^d$.

What are some nontrivial bounds on $p=\mathbb{P}(S)$, where $S$ is the event $Z_1=\max Z$?

This is motivated by the inequality

$$\mathbb{E}(|Z_1|\mathbf{1_S})\ge \max\bigg(0,p\mathbb{E}|Z_1|-\sqrt{\mu_i^2+\Sigma_{ii}-(\mathbb{E}|Z_1|)^2}\sqrt{p-p^2}\bigg)$$

Let $Z\sim \mathcal{N}(\mu, \Sigma)$ be a Gaussian random vector in $\mathbb{R}^d$.

What are some nontrivial bounds on $p=\mathbb{P}(S)$, where $S$ is the event $Z_1=\max_i Z_i$?

This is motivated by the inequality

$$\mathbb{E}(|Z_1|\mathbf{1_S})\ge \max\bigg(0,p\mathbb{E}|Z_1|-\sqrt{\mu_i^2+\Sigma_{ii}-(\mathbb{E}|Z_1|)^2}\sqrt{p-p^2}\bigg)$$

added 20 characters in body
Source Link

Let $Z\sim \mathcal{N}(\mu, \Sigma)$ be a Gaussian random vector in $\mathbb{R}^d$.

What are some nontrivial bounds on $p=\mathbb{P}(S)$, where $S$ is the event $Z_1=\max Z$?

This is motivated by the inequality

$$\mathbb{E}(|Z_1|\mathbf{1_S})\ge \max\bigg(0,p\mathbb{E}|Z_1|-\sqrt{\mu_i^2+\Sigma_{ii}}\sqrt{p-p^2}\bigg)$$$$\mathbb{E}(|Z_1|\mathbf{1_S})\ge \max\bigg(0,p\mathbb{E}|Z_1|-\sqrt{\mu_i^2+\Sigma_{ii}-(\mathbb{E}|Z_1|)^2}\sqrt{p-p^2}\bigg)$$

Let $Z\sim \mathcal{N}(\mu, \Sigma)$ be a Gaussian random vector in $\mathbb{R}^d$.

What are some nontrivial bounds on $p=\mathbb{P}(S)$, where $S$ is the event $Z_1=\max Z$?

This is motivated by the inequality

$$\mathbb{E}(|Z_1|\mathbf{1_S})\ge \max\bigg(0,p\mathbb{E}|Z_1|-\sqrt{\mu_i^2+\Sigma_{ii}}\sqrt{p-p^2}\bigg)$$

Let $Z\sim \mathcal{N}(\mu, \Sigma)$ be a Gaussian random vector in $\mathbb{R}^d$.

What are some nontrivial bounds on $p=\mathbb{P}(S)$, where $S$ is the event $Z_1=\max Z$?

This is motivated by the inequality

$$\mathbb{E}(|Z_1|\mathbf{1_S})\ge \max\bigg(0,p\mathbb{E}|Z_1|-\sqrt{\mu_i^2+\Sigma_{ii}-(\mathbb{E}|Z_1|)^2}\sqrt{p-p^2}\bigg)$$

added 14 characters in body
Source Link
Loading
deleted 1127 characters in body
Source Link
user44143
user44143
Loading
added 4 characters in body
Source Link
Loading
added 357 characters in body
Source Link
Loading
added 796 characters in body
Source Link
Loading
added 1 character in body
Source Link
Loading
Source Link
Loading