泊松参数的无偏估计

9

每天的事故数量是带有参数的泊松随机变量，在随机选择的10天中，观察到的事故数量为1,0,1,1,2,0,2,0,0,1，将是的无偏估计？ $\lambda$ $e^{\lambda}$

我想用这种方式来尝试：我们知道，，但。那么，所需的无偏估计量是多少？ $E(\bar{x})=\lambda=0.8$ $E(e^{\bar{x}})\neq\ e^{\lambda}$

self-study estimation poisson-distribution

— 普里扬卡
source

9

如果，则用于。很难计算 $X\sim \text{Pois}(\lambda)$ $P(X = k) = \lambda^ke^{-\lambda}/k!$ $k\geq 0$

但更容易计算，其中：

E [X^{n}] = \sum_{k \geq 0} k^{n} P (X = k),

$E[X^n] = \sum_{k\geq 0} k^n P(X = k)\text{,}$

E [X^{\underline{n}}]

$E[X^{\underline{n}}]$

X^{\underline{n}} = X (X - 1) \dots (X - n + 1)

$X^{\underline{n}} = X(X - 1)\cdots (X - n + 1)$

E [X^{\underline{n}}] = λ^{n} .

$E[X^\underline{n}]=\lambda^n\text{.}$ 您可以自己证明这一点-这很容易。此外，我让你证明自己执行以下操作：如果

被IID为

，然后

，因此

X_{1}, \dots, X_{N}

$X_1,\cdots, X_N$

Pois (λ)

$\text{Pois}(\lambda)$

U = \sum_{i} X_{i} \sim Pois (N λ)

$U = \sum_i X_i\sim \text{Pois}(N\lambda)$

设

。它遵循

E [U^{\underline{n}}] = (N λ)^{n} = N^{n} λ^{n} and E [U^{\underline{n}} / N^{n}] = λ^{n} .

$E[U^{\underline{n}}] = (N\lambda)^n = N^n \lambda^n\quad\text{and}\quad E[U^\underline{n}/N^n] = \lambda^n\text{.}$

Z_{n} = U^{\underline{n}} / N^{n}

$Z_n = U^{\underline{n}}/N^n$

是您的测量结果的函数，， $Z_n$ $X_1$ $\dots$ $X_N$
， $E[Z_n] = \lambda^n$

由于，我们可以推断出 $e^\lambda = \sum_{n\geq 0}\lambda^n /n!$

E [\sum_{n \geq 0} \frac{Z_{n}}{n!}] = \sum_{n \geq 0} \frac{λ^{n}}{n!} = e^{λ},

$E\left[\sum_{n\geq 0}\frac{Z_n}{n!}\right] =\sum_{n\geq 0} \frac{\lambda^n}{n!} = e^\lambda\text{,}$

W = \sum_{n \geq 0} Z_{n} / n!

$W = \sum_{n\geq 0} Z_n/n!$

E [W] = e^{λ}

$E[W] = e^\lambda$

W

$W$

U \in N_{0}

$U\in \mathbb{N}_0$

U^{\underline{n}} = 0

$U^\underline{n} = 0$

n > U

$n>U$

Z_{n} = 0

$Z_n = 0$

n > U

$n>U$

$\lambda$ $f(\lambda) = \sum_{n\geq 0}a_n\lambda^n$

— 安托万
source

3

$Y=\sum_{i=1}^{10} X_i \sim \text{Pois}(10\lambda)$ $\theta=e^\lambda$

\hat{θ} = e^{\bar{X}} = e^{Y / 10} .

$\begin{equation} \hat\theta = e^{\bar X} = e^{Y/10}. \end{equation}$

Y

$Y$

M_{Y} (t) = e^{10 λ (e^{t} - 1)},

$\begin{equation} M_Y(t)=e^{10\lambda(e^t - 1)}, \end{equation}$

E (\hat{θ}) = E (e^{\frac{1}{10} Y}) = M_{Y} (\frac{1}{10}) = e^{10 λ (e^{1 / 10} - 1)} = θ^{10 (e^{1 / 10} - 1)},

$\begin{equation} E(\hat\theta) = E(e^{\frac1{10}Y}) = M_Y(\frac1{10}) = e^{10\lambda(e^{1/10} - 1)} = \theta^{10(e^{1/10}-1)}, \end{equation}$

\hat{θ}

$\hat\theta$

θ^{*} = e^{a Y},

$\begin{equation} \theta^* = e^{aY}, \end{equation}$

a

$a$

Y

$Y$

E (θ^{*}) = e^{10 λ (e^{a} - 1)} = θ^{10 (e^{a} - 1)},

$\begin{equation} E(\theta^*) = e^{10\lambda(e^a - 1)} = \theta^{10(e^a-1)}, \end{equation}$

10 (e^{a} - 1) = 1

$10(e^a - 1) = 1$

a = \ln \frac{11}{10}

$a=\ln\frac{11}{10}$

θ^{*} = (\frac{11}{10})^{Y}

$\theta^* = (\frac{11}{10})^Y$

θ = e^{λ}

$\theta=e^\lambda$

$Y$ $\lambda$ $\theta^*$ $Y$ $e^\lambda$

— 贾勒·图夫托
source