The Power and Sample Size Application

The description is used to identify this particular project in the Open and Delete windows. Type a description for your project in the Project: text box.

For this example, change the description to Experimental blood pressure drug with two groups, as shown in Figure 90.35.

For the two-sample t test analysis, you can choose to solve for power, sample size per group, or total sample size. Specify the desired quantity type on the Solve For tab.

Click the Solve For tab and select the Power option as shown in Figure 90.35. For information about solving for sample size, see the section Solving for Sample Size.

Click the Distribution tab to select a distribution option that specifies the underlying distribution for the test statistic, as shown in Figure 90.36.

Figure 90.36: Distribution Tab

For this example, you are interested in means rather than mean ratios, so select the Normal option.

Click the Hypothesis tab to select the type of test; see Figure 90.37.

Figure 90.37: Hypothesis Tab

You can choose either a one- or two-sided test. If you do not know the direction of the effect (that is, whether it is positive or negative), the two-sided test is appropriate. If you know the effect’s direction, the one-sided test is appropriate. For the one-sided test, the alternative hypothesis is assumed to be in the same direction as the effect. If you specify a one-sided test and the effect is in the unexpected direction, the results of the analysis are invalid.

The One-sided test option assumes that you know the correct direction of the test. Select the Lower one-sided test and Upper one-sided test options to explicitly indicate the direction of the one-sided test.

Because you are interested only in whether the experimental drug lowers blood pressure, select the One-sided test option on the Hypothesis tab.

Click the Test tab to select either the pooled t test or the Satterthwaite t test.

Figure 90.38: Test Tab

With the independent variances that the example uses, select Pooled t test option. The Satterthwaite t test is used with unequal variances; it is available only with the normal distribution.

Click the Alpha tab to specify one or more significance levels, as shown in Figure 90.39.

Figure 90.39: Alpha Tab

Alpha is the significance level (that is, the probability of falsely rejecting the null hypothesis). If you frequently use the same values for alpha, set them as defaults in the Preferences window. See the section Setting Preferences for more information about setting preferences.

Type the desired significance level of 0.05 in the first cell of the Alpha table (if it is not already the default value).

Click the Means tab to select one of four possible ways to enter the means and the null mean difference, as shown in Figure 90.40.

Figure 90.40: Means Tab

Select one of the following forms from the Select A Form list. The four available forms are:

For this analysis, you can enter the means for the two groups either individually or as a difference. If your null mean difference is not zero, enter that value in the Null Mean table. (The Null Mean table is displayed only for the Group means, Null Difference and Difference between means, Null difference forms.)

For this example, a null mean difference of 0 is reasonable, so select the Group means form from the list, as shown in Figure 90.40. Enter the control mean of 132 in the first row of the first column and the experimental mean of 120 in the first row of the second column.

Click the Standard Deviation tab to enter the standard deviation for the two groups. It is assumed to be equal for both groups.

For the example, enter a single value of 15, as shown in Figure 90.41.

Figure 90.41: Standard Deviation Tab

Click the Sample Size tab to select one of three possible ways to enter the sample sizes, as shown in Figure 90.40.

Select one of the following forms from the Select A Form list:

Examine the alternatives by clicking the Select a form down arrow. For this example, select the Sample size per group form. You want to examine a curve of powers in the power by sample size graph, so enter the values 20, 25, and 30 in the Sample Size table, as shown in Figure 90.42. If you need to add more rows to the table, add them by clicking the button beneath the table.

Figure 90.42: Sample Size Tab

Table 90.2 contains the values of the input parameters for the example.

Click the Results tab to request desired results. Summary table and power by sample size graph options are available.

For the example, select the Create summary table and Create power by sample size graph check boxes.

Click Calculate to perform the analysis. If there are no errors in the input values, the View Results page appears. If there are errors in the input parameter values, you are prompted to correct them.

Click the Summary Table tab to view a table that includes the values of the input parameters and the computed quantity (in this example, power). See Figure 90.43.

Figure 90.43: Results Page with Summary Table

The table consists of two subtables: the Fixed Scenario Elements table that contains the input parameters that have only one value for the analysis, and the Computed Power table that contains the input parameters that have more than one value for the analysis and the corresponding power. Only the N per group parameter appears in the Computed Power table; all of the other input parameters have a single value. The computed power for a sample size per group of 25 is 0.874. Thus, you have a probability of 0.87 that the study will find the expected result if the assumptions and conjectured values are correct.

Click the Graph tab to view a power by sample size graph that displays power on the vertical axis and sample size per group on the horizontal axis. See Figure 90.44.

Figure 90.44: Power by Sample Size Graph

The range of values for the horizontal axis is 20 to 30, which were the smallest and largest values, respectively, that you entered on the Sample Size tab. You can customize the graph by specifying the values for the sample size axis (see the section Customizing Graphs).

Click the Narratives tab to create and display a sentence- or paragraph-length text summary of the input parameter values and the computed quantity for combinations of the input parameter values; see Figure 90.45.

Figure 90.45: Narrative Selector and Display

To create a narrative, selected the desired scenario (row) in the narrative selector table at the bottom of the Narratives tab.

In this example, select the narrative for the sample size per group of 20, which yields a power of 0.799. The following text summary is displayed:

To create other narratives, select the desired rows in the narrative selector table. If you also select the second row for the sample size of 25, another text summary is displayed below the first one:

To change some values of the analysis and rerun it, select the Edit Properties page, change the desired properties, and click the Calculate button again.