91Ó°ÊÓ

When conducting a hypothesis test, a Type I error happens when we falsely reject the null hypothesis, denoted as \(H_0\), even though it is actually true. This error is represented by the Greek letter \(\alpha\), which is also called the "significance level" of the test.

In our hypothesis test scenario, we reject \(H_0\) if our test statistic \(z_0\) is either greater than a critical value \(z_\epsilon\) or less than \(-z_{\alpha-\epsilon}\). These critical values are determined based on the tails of the normal distribution.

The areas in these tails represent the probability that \(z_0\) will fall into these critical regions when \(H_0\) is true. Specifically, the probability \(P(z_0 > z_\epsilon)\) is \(\epsilon\), and \(P(z_0 < -z_{\alpha-\epsilon})\) is \(\alpha - \epsilon\). By adding these probabilities together, we get \(\epsilon + (\alpha - \epsilon) = \alpha\). This means the Type I error rate is exactly \(\alpha\).

While a Type I error involves incorrectly rejecting \(H_0\), a Type II error occurs when we fail to reject \(H_0\) even though the alternative hypothesis \(H_1\) is true. This error is denoted by \(\beta\).

In our test, \(H_1\) is valid when the true mean \(\mu_1\) is different from \(\mu_0\) by some amount \(\delta\). When \(\mu_1 = \mu_0 + \delta\), the distribution of the test statistic shifts. This shift affects the region where we expect the test statistic to fall under \(H_1\).

The probability \(\beta\) describes how likely it is that \(z_0\) stays between \(-z_{\alpha-\epsilon}\) and \(z_\epsilon\) under \(H_1\). The mathematical expression for \(\beta\) includes compensating for this "shift" by \(\delta\), and is given by: \[\beta = P\left(-z_{\alpha-\epsilon} + \frac{\delta \sqrt{n}}{\sigma} < \frac{X - \mu_1}{\sigma/\sqrt{n}} < z_\epsilon + \frac{\delta \sqrt{n}}{\sigma}\right)\]. This expression shows the probability of making a Type II error.

The normal distribution is a continuous probability distribution often known as the bell curve due to its shape. It is significant in hypothesis testing because many test statistics, including our \(z_0\), follow this distribution under the null hypothesis.

In our hypothesis test example, we assume the underlying population distribution is normal. This assumption allows us to compute probabilities and critical values accurately. The standard normal distribution has a mean of 0 and a standard deviation of 1, which we use to find critical values like \(z_\epsilon\) and \(-z_{\alpha-\epsilon}\).

Understanding the properties of the normal distribution helps us to interpret the significance level \(\alpha\) and how it corresponds to the tails of the distribution. Whether we're assessing potential Type I errors or measuring the power of a test against Type II errors, the normal distribution is integral to analyzing the behavior of data in statistical tests.

A test statistic is a standardized value that helps to determine if we should reject the null hypothesis \(H_0\). In our example, the test statistic \(z_0\) measures how far the sample mean is from the hypothesized population mean \(\mu_0\), considering the standard deviation \(\sigma\) and the sample size \(n\).

The formula to calculate \(z_0\) is: \[z_0 = \frac{X - \mu_0}{\sigma/\sqrt{n}}\]. This value tells us the position of our sample mean within the context of the standard normal distribution. Based on where \(z_0\) falls relative to our critical points \(z_\epsilon\) and \(-z_{\alpha-\epsilon}\), we decide whether to reject \(H_0\).

Test statistics are pivotal because they transform sample data into a framework where we can apply probability. This transformation lets us make objective decisions about hypotheses based on pre-determined significance levels, helping to minimize errors in hypothesis testing. By comparing \(z_0\) with critical values, we systematically decide the course of action based on evidence rather than intuition.