91Ó°ÊÓ

In statistical hypothesis testing, the null hypothesis (denoted as \( H_0 \)) is a statement that proposes no effect or no difference — it suggests the status quo or a baseline assumption. It is the hypothesis that researchers aim to test against. In the context of the taste testing example provided, the null hypothesis would be: "The store brand and the national brand have an equal chance of being preferred, i.e., the store brand is not better than the national brand." This translates mathematically to \( P = 0.5 \), where \( P \) is the probability that the store brand is judged as better.

When conducting hypothesis testing, the goal is typically to challenge the null hypothesis in hopes of finding new evidence (alternative hypothesis). If evidence strongly refutes \( H_0 \), researchers may discard it in favor of the alternative hypothesis. It is crucial to start with a clear null hypothesis since all computational aspects of hypothesis testing, such as test choice and significance level determination, rely on it.

The alternative hypothesis (denoted as \( H_1 \)) is what the researcher wants to prove. It suggests a new outcome or effect — a change from the status quo represented by the null hypothesis. In the context of the provided exercise, the alternative hypothesis is that the store brand is judged as better than the national brand more than 50% of the time, represented mathematically as \( P > 0.5 \).

This hypothesis indicates that the store brand is potentially better in taste compared to the national brand. When testing hypotheses, the null and alternative are contrasting, and only one can be favored based on the evidence. Researchers look for enough statistical evidence that would lead them to believe that the alternative hypothesis could be true, resulting in rejection of the null hypothesis. Importantly, this specific test is one-tailed because we are testing the direction of the effect — whether the store brand is "better than" rather than simply "different from" the national brand.

A binomial test is a statistical test used to determine if there is a significant difference between the observed proportion of successes in a binary outcome and a theoretical expected proportion. It is especially handy when working with small binary data sets, as seen in our exercise involving 35 taste test comparisons.

In this scenario, we are assessing whether the store brand is judged as better than the national brand in more than half of the cases. Given the test involves "successes" and "failures" (i.e., store brand preferred or not), and considering the small sample size, the binomial test is suitable.

Using the binomial distribution, we can calculate the probability of observing a certain number of successes. Here, we check the probability that the store brand is preferred at least six times out of 35, given the null hypothesis of a 50% chance \((P = 0.5)\). If this probability is very low (below a pre-determined significance level), it suggests that our observation (store brand being better in this case) is unlikely under the null hypothesis.

In hypothesis testing, the significance level \( \alpha \) is the threshold used to determine whether a hypothesis test's results are statistically significant. It is the probability of rejecting the null hypothesis when it is, in fact, true (a type I error). For example, a significance level of \( \alpha = 0.01 \) means there is a 1% risk of concluding that there is an effect when there is none.

In the taste test exercise, a significance level of 0.01 was chosen, indicating very strict criteria for accepting the alternative hypothesis. If the calculated probability (from our binomial test) is less than this significance level, we reject the null hypothesis in favor of the alternative hypothesis. This decision-making rule ensures that we only reject the null hypothesis when there is strong evidence against it, protecting us from incorrect conclusions.

A lower significance level means the test is more selective about claiming that findings are statistically significant. This mitigates the risk of false positives, ensuring the results are more reliable.