91Ó°ÊÓ

In any statistical test, determining whether results are statistically significant is a key aspect. When we say that results are statistically significant, it means that they are unlikely to have occurred by random chance alone. In the context of our exercise, we are examining the discrepancy in response rates to resumes with different name types.

Determining statistical significance involves calculating a test statistic and comparing it to a critical value. A result is deemed significant if the observed test statistic falls into a region that suggests the null hypothesis should be rejected. This region is often determined by a chosen significance level (usually \( \alpha = 0.05 \)).

• If the test result is statistically significant, it suggests that the factors we are examining have a real, consistent effect.
• If not significant, the observed differences could just be due to chance.

Statistical significance is important because it helps infer whether an observed effect is meaningful beyond mere chance.

The null hypothesis, denoted as \( H_0 \), serves as the starting point in hypothesis testing. It represents a statement of no effect or no difference. In our example problem, the null hypothesis is that the probability of receiving a response is the same for resumes with "white" sounding names and "black" sounding names.

This hypothesis serves as a benchmark for comparison. When conducting a statistical test, if we obtain evidence against \( H_0 \), it suggests that an alternative reality, described by the alternative hypothesis, may better explain our observations.

• Rejection or failure to reject the null hypothesis informs decision making.
• Null hypotheses are assumed true unless the evidence strongly indicates otherwise.

Despite its central role, it's crucial to understand that suspecting the null hypothesis doesn't prove any specific alternative true. It simply indicates a need to explore other potential explanations.

The alternative hypothesis, denoted \( H_1 \) or \( H_a \), is the counterpart to the null hypothesis. It reflects the statement we're looking to find evidence for in a hypothesis test. In our scenario, the alternative hypothesis suggests that resumes with "black" names are less likely to receive a response than those with "white" names.

This hypothesis is directional in this case because it specifically proposes that one group has a lower rate than the other. While the null hypothesis assumes no difference, the alternative hypothesis highlights what we aim to prove statistically.

• If the test results lead to rejecting \( H_0 \), the alternative hypothesis gains support.
• It is important to clearly define \( H_1 \) before performing the test to avoid biases.

The consequence is that accepting the alternative hypothesis provides justification for considering deeper examination or potential policy changes.

A test statistic is a standard value derived from sample data and used to make an inference about a population parameter. For comparing two proportions, as in our exercise, a z-test is typically employed. The test statistic z reflects how far away our sample result is from what we would expect under the null hypothesis.

In our example, the proportion of responses for "white" names \( (p_w) \) and "black" names \( (p_b) \) are compared using this z-statistic. The formula for the z-test here is:

\[ z = \frac{(p_w - p_b)}{\sqrt{p(1-p)(\frac{1}{n_1}+\frac{1}{n_2})}} \]

where \( p \) is the pooled sample proportion.

This z-value assists in determining whether the observed difference is due to chance. It is compared to a critical value from the standard normal distribution:

• If the z-value exceeds a certain threshold (critical value), \( H_0 \) is rejected.
• The decision is usually based on pre-specified significance levels.

The test statistic is a fundamental building block for evaluating hypotheses and drawing conclusions in statistics.