1) Critical Region and Most Powerful Critical Region
Critical Region:
- The critical region (rejection region) is the range of values of a test statistic where the null
hypothesis H0 is rejected.
- It is determined based on the selected test statistic and the level of significance (α).
- The boundary separating the critical region and acceptance region is called the critical
value.
- If the observed test statistic exceeds this boundary, it indicates that the sample data is highly
inconsistent with H0.
- Example: In a right-tailed Z-test, critical region is Z > Zα (like 1.645 for α = 0.05).
Most Powerful Critical Region:
- A critical region is called most powerful if, among all tests with the same level of significance
α, it gives the maximum probability of correctly rejecting H0 when H1 is true.
- The power of a test is P(reject H0 | H1 true). A most powerful test maximizes this probability.
- Based on the Neyman–Pearson Lemma, the most powerful test for testing two simple
hypotheses is the likelihood ratio test.
- Such a region provides the strongest decision rule against H0 for a fixed risk (α).
2) Level of Significance and Power of Test
Level of Significance (α):
- It is the probability of committing a Type I Error, i.e., rejecting a true null hypothesis.
- It represents the maximum allowable risk of making a false rejection.
- Common values: 0.01, 0.05, 0.10.
- Lower α → stricter test → lower chance of Type I error, but higher chance of Type II error.
- Example: α = 0.05 means we are willing to take 5% risk of rejecting H0 when it is actually
true.
Power of a Test:
- The power of a test is the probability of correctly rejecting H0 when H1 is true.
- Power = 1 – β, where β = probability of Type II Error.
- A more powerful test is more sensitive in detecting actual differences or effects.
Factors affecting power:
1. Sample size (n): Larger n → higher power.
2. Effect size: Larger difference between true population value and H0 → higher power.
3. Variance: Lower variance → higher power.
4. Level of significance (α): Higher α increases power but increases Type I Error.
- Researchers often aim for power ≥ 0.80.
3) Population and Sample
Population:
- The complete set of individuals, items, or measurements under study.
- Can be finite or infinite.
- Example: All students in a university, all manufactured bolts in a factory.
- Parameters such as population mean (µ) and population variance (σ²) describe it.
Sample:
- A subset of the population selected for analysis.
- Used when studying the entire population is impractical or expensive.
- Corresponding statistical measures are sample mean (x■) and sample variance (s²).
- Good sampling must be unbiased and representative of the population.
4) Type I and Type II Errors
Type I Error (α):
- Occurs when a true null hypothesis is rejected.
- Known as a "false positive".
- Probability of Type I Error = α.
- Example: Concluding a medicine works when it actually does not.
Type II Error (β):
- Occurs when a false null hypothesis is not rejected.
- Known as a "false negative".
- Probability of Type II Error = β.
- Example: Concluding a medicine does not work when it actually works.
Relation: Decreasing α increases β and vice versa.
5) Critical Region (Detailed)
- The critical region is the set of values of a test statistic that leads to the rejection of H0.
- Determined by α and the sampling distribution of the test statistic.
Types of Critical Regions:
1. Right-tailed: Reject H0 if statistic > critical value.
2. Left-tailed: Reject H0 if statistic < critical value.
3. Two-tailed: Reject H0 if statistic is too high or too low.
- Example: For a two-tailed Z-test at α = 0.05, critical region is Z < -1.96 or Z > 1.96.
6) Power of Test (Detailed)
- Power is the ability of a test to detect a false null hypothesis.
- Higher power means the test is more efficient and reliable.
Graphical Interpretation:
- It is the area under the sampling distribution of the test statistic under H1 that lies inside the
rejection region.
Improving power:
- Increase sample size.
- Increase α (more risk allowed).
- Reduce variability through better measurement techniques.
- Choose a more suitable test statistic.