9.7 Probability of a Type II Error and the Power of a Hypothesis Test

OBJECTIVES By the end of this section, I will be able to …

  1. Calculate the probability of a Type II error for a test for .
  2. Compute the power of a test for and construct a power curve.

1 Probability of a Type II error

In Section 9.1, we defined a Type II error as follows:

For example, the criminal trial scenario on page 494 had the following hypotheses:

565

In this case, a Type II error was to find the defendant not guilty (not reject ) when in reality he did commit the crime ( is false). In this section, we learn how to calculate the probability of making a Type II error for a test for , called (beta), and to use the value of to compute the power of a test for .

Calculating , the Probability of a Type II Error

Use the following steps to calculate , the probability of a Type II error.

  • Step 1 Recall that divides the critical region from the noncritical region. Let be the value of the sample mean associated with . The following table shows how to calculate for the three forms of the hypothesis test.

    Form of test Value of
    Right-tailed
    Left-tailed
    Two-tailed

    Here, is the hypothesized value of the population mean, σ is the population standard deviation, and is the sample size.

  • Step 2 Let represent a particular value for the population mean chosen from the values indicated in the alternative hypothesis . Draw a normal curve centered at , with the value or values of from Step 1 indicated (see Example 34).
  • Step 3 Calculate for the particular chosen using the following table.

Form of test β = probability of Type II error
Right-tailed The area under the normal curve drawn in Step 2 to the left of .
Left-tailed The area under the normal curve drawn in Step 2 to the right of .
Two-tailed The area under the normal curve drawn in Step 2 between and .

Let us illustrate the steps for calculating , the probability of a Type II error, using an example.

EXAMPLE 34 Calculating , the probability of a Type II error

ATM network operator Star Systems of San Diego reported that active users of debit cards used them an average of 11 times per month. Suppose we are interested in testing whether people use debit cards on average more than 11 times per month, using level of significance . The hypotheses are

566

where represents the population mean debit card usage per month. Suppose we have , and , and from Table 4 (page 500) we have .

  1. State what a Type II error would be in this case.
  2. Let . That is, suppose the population mean debit card usage is actually 13 times per month. Calculate , the probability of making a Type II error when .

Solution

  1. We make a Type II error when we do not reject when is false. In this case, a Type II error would be to conclude that the population mean debit card usage was 11 times per month when in actuality it was more than 11 times per month.
  2. We follow the steps for calculating .
    • Step 1 We have a right-tailed test, so that

    • Step 2 Figure 49 shows the normal curve centered at , with labeled.

      image
      Figure 9.60: FIGURE 49 probability of Type II error.
    • Step 3 The right-tailed test tells us that equals the area under the normal curve drawn in Step 2 to the left of . This is the shaded area in Figure 49. Area represents probability, so we have

      Standardizing with , and :

      This is a Case 1 problem from Table 8 of Chapter 6 on page 355.

      Thus, . This represents the probability of making a Type II error, that is, of not rejecting the hypothesis that the population mean debit card usage is 11 times per month when in actuality it is 13 times per month.

NOW YOU CAN DO

Exercises 5–16.

567

2 Power of a Hypothesis Test

It is a correct decision to reject the null hypothesis when the null hypothesis is false. The probability of making this type of correct decision is called the power of the test.

Power of a Hypothesis Test

The power of a hypothesis test is the probability of rejecting the null hypothesis when the null hypothesis is false. Power is calculated as

EXAMPLE 35 Power of a hypothesis test

Calculate the power, for the particular alternative value of the mean, of the hypothesis test in Example 34.

Solution

The probability of a Type II error was found in Example 34 to be . Thus, the power of the hypothesis test is

The probability of correctly rejecting the null hypothesis is 0.9525.

NOW YOU CAN DO

Exercises 17–28.

image What if Scenario: Type II Errors and Power of the Test

Suppose that we have the same hypothesis test from Example 34 and the same value . Now, what if we decrease such that it is less than 13 but still larger than 12.165? Describe what will happen to the following, and why:

  1. The probability of a Type II error,
  2. The power of the test,

Solution

  1. Consider Figure 50. The distribution of sample means remains centered at , so that a smaller will “slide” the normal curve toward the value of . This results in a larger area to the left of 12.165, as you can see by comparing Figure 50 with Figure 49. Therefore, a smaller leads to an increase in the probability of a Type II error, .

  2. As increases, decreases. Therefore, a smaller leads to a decrease in the power of the test.
    image
    Figure 9.61: FIGURE 50 Smaller leads to an increase in .

568

A power curve plots the values for the power of the test versus the values of .

EXAMPLE 36 Power curve

  1. Calculate the power of the hypothesis test from Example 34 for the following values of .
  2. Construct the power curve by graphing the values for the power of the test on the vertical axis against the values of on the horizontal axis.

Solution

  1. We have , and . The calculations are provided in the following table.
    Probability of Type II error: Power of the test:
    11.0
    11.5
    12.0
    12.165
    12.5
    13.5
  2. Figure 51 represents a power curve, because it plots the values for the power of the test on the vertical axis against the values of on the horizontal axis. Note that, as moves farther away from the hypothesized mean , the power of the test increases. This is because it is more likely that the null hypothesis will be correctly rejected as the actual value of the mean gets farther away from the hypothesized value .

    image
    Figure 9.62: FIGURE 51 Power curve.

For completeness, we include the power for from Example 35 in this power curve.

NOW YOU CAN DO

Exercises 29 and 30.