0
$\begingroup$

I have some trouble getting the standard deviation of a stratified sample.

I have a sample of 1600 individuals that are divided into three groups (strata), with their sample mean $y$, which is the groups average income, $s_k$ is the standard deviation, and $n_k$ is the number of indivdiuals in the strata/group.

What is the standard deviation of the population?

enter image description here

$\endgroup$
1
$\begingroup$

Answer is given here. Compute population mean ($\bar y$). Then
Population sample deviation = $\sqrt(\sum_k n_k s_k^2 + \sum_k n_k(\bar y_k-\bar y)^2)$.

$\endgroup$
0
$\begingroup$

You have 3 stratas with sizes $n_1$, $n_2$, $n_3$. Suppose, observations $y_1, \dots, y_{n_1}$ belong to the first strata, observations $y_{n_1 + 1}, \dots, y_{n_2}$ belong to the second strata, observations $y_{n_2 + 1}, \dots, y_{n_3}$ belong to the third strata. Mean values of stratas are $\overline{y}_1, \overline{y}_2, \overline{y}_3$ correspondingly, in the same way we define $n_1, n_2, n_3$ and $s_1, s_2, s_3$. Also, $n = n_1 + n_2 + n_3$

To get the standard deviation we need to evaluate $\frac{1}{n} \sum_{i = 1}^{n} (y_i - \overline{y})^2$.

Given $n_i$, $\overline{y}_i$ it is easy to calculate $\overline{y} = \frac{1}{n} \sum_{i = 1}^n y_i = \frac{1}{n} \sum_{i = 1}^3 n_i \overline{y}_i$. Now let us calculate a part of the sum up to the $n_1$-th term $\sum_{i = 1}^{n_1} (y_i - \overline{y})^2$: $$ \sum_{i = 1}^{n_1} (y_i - \overline{y})^2 = \sum_{i = 1}^{n_1} (y_i - \overline{y}_1 + \overline{y}_1 - \overline{y})^2 = \\ \sum_{i = 1}^{n_1} (y_i - \overline{y}_1)^2 + n_1 (\overline{y}_1 - \overline{y})^2 + 2 \sum_{i = 1}^{n_1} (y_i - \overline{y}_1) (\overline{y}_1 - \overline{y}) = \\ \sum_{i = 1}^{n_1} (y_i - \overline{y}_1)^2 + n_1 (\overline{y}_1 - \overline{y})^2 = n_1 s_1^2 + n_1 (\overline{y}_1 - \overline{y})^2. $$

For other terms in the sum we do the same thing and get the final answer.

The key idea here is to add a term $0 = (\overline{y}_1 - \overline{y}_1)$ and then open the parenthesis.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.