Tests of Significance - Study Notes
1. Inferential statistics
- Descriptive statistics (Class 11) summarises data; inferential statistics uses random samples to draw valid inferences about a population in terms of probability statements.
- Examining the whole population costs time and money, so a part of it, the sample, is examined.
2. Parameter and statistic
- Parameter: a quantitative characteristic that indexes the population distribution, e.g. \( \mu \), \( \sigma \), \( P \). Denoted by Greek letters; assumed unknown in inference.
- Random sample: realisations \( (X_1, \ldots, X_n) \) made under independent and identical conditions (iid).
- Statistic: any quantity calculated from the random sample, e.g. \( \bar{X} \), \( S \), \( p \); denoted by Roman letters. A statistic is itself a random variable and has a probability distribution.
- Sample standard deviation: \( S = \sqrt{\dfrac{1}{n-1}\sum (X_i - \bar{X})^2} \); the divisor is \( n - 1 \).
- The set of all possible observed samples is the sample space S.
| Measure | Parameter | Statistic | Value for a sample |
|---|---|---|---|
| Mean | \( \mu \) | \( \bar{X} \) | \( \bar{x} \) |
| Standard deviation | \( \sigma \) | S | s |
| Proportion | P | p | \( p_0 \) |
3. Sampling distribution
- The probability distribution of a statistic is its sampling distribution.
- From a population of size N, samples of size n number \( N^n \) with replacement and \( ^NC_n \) without replacement.
- Example 1.1: population 4, 8, 12, 16, n = 2, with replacement: 16 samples; the sample means 4 to 16 have probabilities 1, 2, 3, 4, 3, 2, 1 out of 16.
- Sampling with replacement gives a random sample; without replacement the independence of \( X_1, \ldots, X_n \) is violated, so it is not a random sample.
- A sample of size 30 or more is a large sample; less than 30 is a small sample.
- If \( X_i \) are iid \( N(\mu, \sigma^2) \), the sample total is \( N(n\mu, n\sigma^2) \) and \( \bar{X} \) is \( N(\mu, \sigma^2/n) \).
4. Standard error
- The standard deviation of the sampling distribution of a statistic is its standard error (SE).
- \( SE(\bar{X}) = \sigma/\sqrt{n} \); \( E(\bar{X}) = \mu \).
- Example 1.2: \( \mu = 10 \), \( \sigma^2 = 20 \), n = 2, so \( V(\bar{X}) = 10 \) and \( SE = \sqrt{10} \).
- \( SE(p) = \sqrt{PQ/n} \); \( SE(\bar{X} - \bar{Y}) = \sqrt{\sigma_X^2/m + \sigma_Y^2/n} \) or \( \sigma\sqrt{1/m + 1/n} \); \( SE(p_X - p_Y) = \sqrt{P_XQ_X/m + P_YQ_Y/n} \), or \( \sqrt{\hat{p}\hat{q}(1/m + 1/n)} \) when \( P_X, P_Y \) are unknown.
5. Hypotheses
- A hypothesis is a statement on the population or on the values of its unknown parameters.
- Null hypothesis \( H_0 \): the hypothesis actually tested for possible rejection; generally one of no difference in comparisons, or that two attributes are independent in the Chi-square test.
- Alternative hypothesis \( H_1 \): contradicts \( H_0 \). For \( H_0: \mu = \mu_0 \): \( \mu \neq \mu_0 \) (two-sided), \( \mu > \mu_0 \) (one-sided right), \( \mu < \mu_0 \) (one-sided left).
6. Errors
| \( H_0 \) true | \( H_0 \) false | |
|---|---|---|
| Reject \( H_0 \) | Type I error | Correct decision |
| Do not reject \( H_0 \) | Correct decision | Type II error |
7. Level of significance, critical region, critical value
- Level of significance \( \alpha \): the maximum probability of type I error we tolerate; usually 5% or 1%, fixed before sampling.
- Critical region: subset of the sample space whose elements lead to rejection of \( H_0 \); its complement is the acceptance region.
- Test statistic: a function of statistics and known parameter values used to decide on \( H_0 \).
- Critical value: the value separating the rejection and acceptance regions, found from the sampling distribution under \( H_0 \).
8. One-tailed and two-tailed tests
- Right-tailed: reject when \( z_0 > z_{\alpha} \). Left-tailed: reject when \( z_0 < -z_{\alpha} \). Both are one-tailed tests.
- Two-tailed: reject when \( |z_0| \geq z_{\alpha/2} \).
9. General procedure (7 steps)
- Describe the population and frame \( H_0 \) and \( H_1 \).
- Describe the sample.
- Specify \( \alpha \).
- Specify the test statistic under \( H_0 \).
- Calculate its value.
- Find the critical value from the table.
- Decide by the rejection rule.
10. Large sample tests
| Test | Statistic |
|---|---|
| Mean, \( \sigma \) known | \( z_0 = \dfrac{\bar{x} - \mu_0}{\sigma/\sqrt{n}} \) |
| Mean, \( \sigma \) unknown | \( z_0 = \dfrac{\bar{x} - \mu_0}{s/\sqrt{n}} \) |
| Two means, variances known | \( z_0 = \dfrac{\bar{x} - \bar{y}}{\sqrt{\sigma_X^2/m + \sigma_Y^2/n}} \) |
| Two means, variances unknown | \( z_0 = \dfrac{\bar{x} - \bar{y}}{\sqrt{s_X^2/m + s_Y^2/n}} \) |
| Proportion | \( z_0 = \dfrac{p - p_0}{\sqrt{p_0q_0/n}} \) |
| Two proportions | \( z_0 = \dfrac{p_X - p_Y}{\sqrt{\hat{p}\hat{q}(1/m + 1/n)}} \) |
- Conditions: \( n \geq 30 \) (both \( m, n \geq 30 \) for two samples); for proportions also \( np > 5 \) and \( n(1-p) > 5 \).
- Book examples: LED bulbs \( z_0 = -3.33 \), reject; cables \( z_0 = 3.87 \), reject; vehicles \( z_0 = 10 \), reject; districts \( z_0 = 8.77 \), reject; boys and girls \( z_0 = -1.75 \), do not reject; tea and coffee \( z_0 = 3.79 \), reject; self-employment \( z_0 = 2.0764 \), reject.
More for this chapter
Book Back Questions18 textbook MCQs · solved
Additional MCQs15 extra MCQs · solved
Practice TestInteractive · instant score
Book Back TestTest yourself on the textbook set
Additional MCQ TestTest yourself on the extra set
Formula SheetAll key formulas