2.1 Introduction

In statistical inference, we assume our data \(X_1, X_2, \dots , X_n\) comes from a distribution \(f(x; \theta )\), but the true value of \(\theta \) is hidden from us. Our goal is to use the sample to construct a “best guess” for the unknown parameter \(\theta \in \Omega \), where \(\Omega \) represents the parameter space.

Definition 2.1.1. A statistic \(T = T(X_1, X_2, \cdots , X_n)\), if a function of the data which does not depend on the unknown parameter \(\theta \).

Example 2.1.2. The sample mean \(\bar {X} = \frac {1}{n}\sum X_i\) is a statistic. While, the quantity \(\frac {\bar {X} - \mu }{\sigma / \sqrt {n}}\) is not a statistic if \(\mu \) or \(\sigma \) are unknown.

Definition 2.1.3. An Estimator is a statistic used to approximate an unknown parameter. It is a random variable denoted by \(\hat {\theta }\).

Definition 2.1.4. An Estimate is the specific numerical value of the estimator obtained from a realized data set \((x_1, x_2, \dots , x_n)\).

Definition 2.1.5. Let \(X_1, X_2, \cdots , X_n\) be a random sample from a population (with a population distribution), then

(a).
the \(k^{\text {th}}\) sample moment about 0 is given by \[M_k' = \frac {1}{n}\sum ^n_{i =1}X^k_i\]
(b).
the \(k^{\text {th}}\) sample moment about the mean \(\overline {X}\) is given by \[M_k = \frac {1}{n}\sum ^n_{i = 1}(X_i - \overline {X})^k.\]

Note 2.1.6.

1.
(a)
\(\mu '_k = E(X^k)\) is the \(k^{\text {th}}\) (population) moment at 0.
(b)
\(\mu _k = E(X - \mu )^k\) is the \(k^{\text {th}}\) (population) moment about the mean \(\mu \).
2.
\(M'_1 = \frac {1}{n}\sum ^n_{i = 1}X_i = \overline {X}, \quad M_1 = 0\)

Theorem 2.1.7. Let \(X_1, X_2, \cdots , X_n\) be a random sample from a population (with a population distribution), then \[E(M'_k) = \mu '_k \quad \quad (\text {if}\quad \mu '_k\quad \text {exists})\] \[\operatorname {Var}(M_k') = \frac {1}{n}\left (\mu '_{2k} - (\mu '_k)^2\right ) \quad (\text {if}\quad \mu '_{2k}\quad \text {exist})\]

i.e. if \(k = 1\), then \[E(\overline {X}) = \mu _1' = E(X) = \mu \] \begin {align*} \operatorname {Var}(\overline {X}) & = \frac {1}{n}\left [\mu _2' - (\mu _1')^2 \right ]\\ & = \frac {1}{n}\left [E(X^2) - (E(X))^2\right ]\\ & = \frac {1}{n}\, \operatorname {Var}(X)\\ & = \frac {\sigma ^2}{n}. \end {align*}

Definition 2.1.8. A statistic \(T = T(X_1,X_2, \cdots , X_n)\) that is used to estimate \(\theta \) or \(\tau (\theta )\), a function of \(\theta \), is called an estimator of \(\tau (\theta )\) and an observed value of the statistic \(t = t(x_1, x_2, \cdots , x_n)\) is called an estimate of \(\tau (\theta )\).

Example 2.1.9. If \(X_1, X_2, \cdots , X_n\) are i.i.d random variables with \(E(X_i) = \mu \) then \(\overline {X}\) is an estimator of \(\mu \) and \(\overline {x}\) is an estimate of \(\mu \).

There are a number of methods of finding estimators of \(\theta \) or \(\tau (\theta )\), some of them include:

1.
method of moments (MME)
2.
maximum likelihood method (MLE)
3.
least squares method (mainly in regression) (LSE)

Questions on this section

Stuck on something here? Ask below and it stays attached to this topic.