91Ó°ÊÓ

In hypothesis testing, the **null hypothesis** ( $$H_0$$ ) is the default claim that no effect or no difference is present, and it often represents the status quo. In our maze example, the null hypothesis is that the proportion of mice completing the maze in less than 18 seconds is equivalent to that observed under normal conditions without any external factors - specifically, that proportion is 0.5.
The null hypothesis serves as a starting point for any statistical test and aims to be tested for rejection. If data sufficiently contradicts the null hypothesis, researchers may consider alternative explanations.

• The null hypothesis often denotes what is generally accepted as true until evidence suggests otherwise.
• In our study, $$H_0: p = 0.5$$ assumes no impact of the noise on maze completion speed.

The **alternative hypothesis** ( $$H_a$$ ) proposes a new effect or a difference contrary to the null hypothesis. In this case, the researcher posits that a loud noise will improve the maze completion time for mice, suggesting a change in the proportion of those completing the maze in less than 18 seconds.
The alternative hypothesis for our maze test is that the proportion of mice that complete the maze in less than 18 seconds is greater than 0.5 due to the noise stimulus. Therefore, our hypothesis takes the form:
$$H_a: p > 0.5$$
This hypothesis challenges the current norm and typically reflects the reason for conducting the experiment.

• The alternative hypothesis is what researchers hope to demonstrate.
• It reflects the direction of the expected effect, such as a "greater than" condition in this mouse study, indicating a positive effect.

**Proportion testing** involves examining whether the proportions derived from our sample data significantly differ from a predefined proportion in the population. This is particularly useful when dealing with binary outcomes, like our mice completing or not completing the maze under the time constraint.
This type of hypothesis test is implemented when data can be expressed in proportions, often using a sample proportion ($$\widehat{p}$$). In our scenario, we're given a sample proportion of 0.7, meaning 70% of the 40 mice completed the maze faster with noise. We then compare this with the hypothesis that the true proportion of speedier maze completion under normal conditions was 0.5.

• Proportion tests help determine if observed differences in sample outcomes reflect true differences in the population.
• They require knowledge of both the population proportion and the sampled proportion, providing insights into potential impacts.

**Statistical significance** is a critical concept in hypothesis testing. It determines whether the observed effect in data is unlikely to have occurred due to random chance, thereby supporting the alternative hypothesis. In essence, it helps to infer if a result from a sample can be generalized to a larger population.
To determine significance, researchers use a threshold known as the "significance level" (often set at 0.05). If the p-value (probability of observing the data assuming the null hypothesis is true) is less than this level, the null hypothesis is rejected.
In our mouse experiment, if testing showed a statistically significant effect, it might imply that noise indeed impacts the mice's maze completion speed. In turn, this would support the researcher’s belief about noise impacting the rate beyond a mere fluke.

• Significance indicates results are meaningful and not just artifacts of random variability.
• Achieving statistical significance provides stronger evidence for accepting the alternative hypothesis.