Skip to main content
MathOverflow

Return to Question

removed capitals from title
Link
YCor
YCor
  • 63.9k
  • 5
  • 187
  • 286

Tight Tail Boundstail bounds for sums of random variables

added 6 characters in body
Source Link
user14097523067
user14097523067
  • 1
  • 1

Let $X_1, X_2, \dots$ be iid uniformly on $[0,1]$. Define $Z_i^a = (X_i - a)^2$$Z_i^{(a)} = (X_i - a)^2$. Let $Y_n = \sum_{k=1}^n Z_k^{1/k}$$Y_n = \sum_{k=1}^n Z_k^{(1/k)}$. I am interested in matching tail bounds for $Y_n$ as $n \to \infty$. In particular, I would like to show that $$C_1e^{-C_2n} \leq P(Y_n < \theta n) \le C_3e^{-C_2n},$$ where $0< \theta < \mathbb{E}[Y_n]/n$ and $C_1, C_2, C_3$ depend only on $\theta$. The difficulty comes from the fact that both bounds have the same constant factor $C_2$ in the exponent. It would also be great to get an expression for $C_2$ in terms of $\theta$.

Let $X_1, X_2, \dots$ be iid uniformly on $[0,1]$. Define $Z_i^a = (X_i - a)^2$. Let $Y_n = \sum_{k=1}^n Z_k^{1/k}$. I am interested in matching tail bounds for $Y_n$ as $n \to \infty$. In particular, I would like to show that $$C_1e^{-C_2n} \leq P(Y_n < \theta n) \le C_3e^{-C_2n},$$ where $0< \theta < \mathbb{E}[Y_n]/n$ and $C_1, C_2, C_3$ depend only on $\theta$. The difficulty comes from the fact that both bounds have the same constant factor $C_2$ in the exponent. It would also be great to get an expression for $C_2$ in terms of $\theta$.

Let $X_1, X_2, \dots$ be iid uniformly on $[0,1]$. Define $Z_i^{(a)} = (X_i - a)^2$. Let $Y_n = \sum_{k=1}^n Z_k^{(1/k)}$. I am interested in matching tail bounds for $Y_n$ as $n \to \infty$. In particular, I would like to show that $$C_1e^{-C_2n} \leq P(Y_n < \theta n) \le C_3e^{-C_2n},$$ where $0< \theta < \mathbb{E}[Y_n]/n$ and $C_1, C_2, C_3$ depend only on $\theta$. The difficulty comes from the fact that both bounds have the same constant factor $C_2$ in the exponent. It would also be great to get an expression for $C_2$ in terms of $\theta$.

Source Link
user14097523067
user14097523067
  • 1
  • 1

Tight Tail Bounds for sums of random variables

Let $X_1, X_2, \dots$ be iid uniformly on $[0,1]$. Define $Z_i^a = (X_i - a)^2$. Let $Y_n = \sum_{k=1}^n Z_k^{1/k}$. I am interested in matching tail bounds for $Y_n$ as $n \to \infty$. In particular, I would like to show that $$C_1e^{-C_2n} \leq P(Y_n < \theta n) \le C_3e^{-C_2n},$$ where $0< \theta < \mathbb{E}[Y_n]/n$ and $C_1, C_2, C_3$ depend only on $\theta$. The difficulty comes from the fact that both bounds have the same constant factor $C_2$ in the exponent. It would also be great to get an expression for $C_2$ in terms of $\theta$.