This procedure can calculate power, sample size and minimum difference for 2 Proportions.
1. Choose ProcessMA > Statistics > Power and Sample Size > 2 Proportions.
2. In Objective, choose what you want to calculate.
· If you want to find sample size(s):
- In Proportion 1 value(s), enter one or more p values, separated by commas(,).
- In Power value(s), enter one or more power values, separated by commas(,).
· If you want to find proportion 1 value(s):
- In Sample size(s), enter one or more sample sizes, separated by commas(,).
- In Power value(s), enter one or more power values, separated by commas(,).
· If you want to find power value(s):
- In Sample size(s), enter one or more sample sizes, separated by commas(,).
- In Proportion 1 value(s), enter one or more p values, separated by commas(,).
3. In Proportion 2, enter the proportion 2 value.
4. In Alternate, select the type of test.
5. In Significance level, enter the desired significance level.
6. Click OK.
Sample size(s): Integer.
Proportion 1 value (s): Numeric; Between 0 to 1.
Power value(s): Numeric; Between 0 to 1; Must be greater than the significance level.
Proportion 2: Numeric; Between 0 to 1.
You want to find out if there is a difference between the proportion of men and proportion of women who support the current government. In the last election, the current government won with a 45% vote. If you survey 500 people, what is the power to detect differences greater than 0.05 between the proportions of men and women at a confidence level of 80% (a = 0.2)?
1. Choose ProcessMA > Statistics > Power and Sample Size > 2 Proportions.
2. In Objective, select Find power value(s).
3. In Sample sizes(s), enter 500.
4. In Proportion 1 value(s), enter 0.4, 0. 5.
5. In Proportion 2, enter 0.45.
6. In Alternate, select Not equals.
7. In Significance level, enter 0.2.
8. Click OK.
Interpretation
For a sample size of 500, you have about 62% chance of detecting a difference of ±0.05 in the proportions of males and females who support the current government.