91Ó°ÊÓ

In hypothesis testing, a **one-tailed test** is used when we want to test if a parameter, such as the population mean (\( \mu \)), is significantly greater than or less than a known value. This type of test helps us determine if there is an effect in a specific direction. For example, you might suspect that a new drug increases blood pressure, so you are only interested in detecting increases.
In a one-tailed test, the null hypothesis (\( H_0 \)) is typically set up like \( H_0: \mu = \mu_0 \). The alternative hypothesis (\( H_1 \)) can then be either \( \mu > \mu_0 \) or \( \mu < \mu_0 \), depending on the direction of interest.
One major feature of a one-tailed test is that the entire level of significance (\( \alpha \)) is concentrated in one tail of the distribution. So, if \( \alpha = 0.01 \), all 1\% would be in one tail, making it easier to reach significance because we are only evaluating one side.
This approach is more sensitive

• Because it's focused on one direction
• There is a greater area for extreme results, hence a lower critical value to cross for significance.

A **two-tailed test** is applied when you are interested in detecting an effect in either direction, whether an increase or a decrease from a known value. This is used when changes in either direction are significant for analysis, such as when any deviation from a standard value affects decision-making.
Here, the null hypothesis remains \( H_0: \mu = \mu_0 \), but the alternative hypothesis becomes \( H_1: \mu eq \mu_0 \), which means testing for any change from the null value, not just an increase or decrease.
The significance level \( \alpha \) for a two-tailed test is split between both tails of the distribution. For a 0.01 significance level, this means each tail assesses an extreme result of 0.005.
This setup is less powerful than a one-tailed test for the same sample and significance level because the critical region is now divided:

• Each tail has a smaller area
• Higher critical values are needed to reject \( H_0 \)

Thus, more extreme test statistics are necessary compared to the one-tailed test.

The **significance level (\( \alpha \))** is a fundamental aspect of hypothesis testing, representing the threshold for determining whether an observed effect is real or merely due to chance. It's the probability of rejecting the null hypothesis \( H_0 \) when it's actually true.
Common significance levels are 0.05, 0.01, and 0.001, implying a 5\%, 1\%, and 0.1\% risk of making a Type I error, respectively. A smaller \( \alpha \) means stricter criteria, requiring stronger evidence to reject \( H_0 \).
In practical terms, the significance level

• Determines how extreme data must be to reject \( H_0 \)
• Helps in specifying critical values

Deciding the value of \( \alpha \) should reflect the context of a study; serious consequences of error demand a smaller \( \alpha \).
For example, in medical research where a false positive might mean approving an ineffective treatment, we might use a smaller \( \alpha \) to reduce the risk of Type I errors.

**Critical values** are the threshold points on the distribution beyond which we consider the results to be statistically significant for the chosen \( \alpha \). They are determined based on the significance level and the type of test used. The critical values help us decide whether to reject the null hypothesis.
In a one-tailed test with \( \alpha = 0.01 \), the critical value could be around 2.33 for the upper tail, or -2.33 for the lower tail, indicating that only about 1\% of the distribution is more extreme.
For a two-tailed test at the same level, critical values change due to the split \( \alpha \):

• The critical values are around ±2.575, as the 1% significance level is divided into two tails.
• Less area on each side means higher criteria for significance.

In both test types, if a test statistic surpasses these critical values, we reject \( H_0 \).
Understanding critical values:

• Aids in interpreting outcomes of hypothesis tests
• Highlights the difference in sensitivity between one-tailed and two-tailed tests

Critical values serve as benchmarks, guiding us through the decision-making process in statistical analysis.