The Associated Press and petside.com survey (2009) found that 63% of dog owners and 53% of cat owners would be likely to give CPR to their pet. Which test would compare these two proportions?

The Associated Press and petside.com survey (2009) found that 63% of dog owners and 53% of cat owners would be likely to give CPR to their pet. Which test would compare these two proportions?

• A. 2-proportion z-test
• B. Fisher's exact test
• C. Paired t-test
• D. ANOVA

Answer: (A)

When you’re comparing two independent groups on a yes/no outcome, you’re looking at two proportions and asking if the difference you observe is real or just due to sampling variation. The best-fitting test is the two-proportion z-test. It evaluates whether the difference between the sample proportions (63% for dog owners vs 53% for cat owners) is large enough to conclude a real difference in the population proportions. It does this by comparing the observed difference to the standard error of that difference under the null hypothesis that the true proportions are equal. Large-sample conditions enable the normal (z) approximation to the binomial, which is why this test is appropriate here. If the counts were very small, Fisher’s exact test or a chi-square test on a 2x2 table might be used instead; the other options—paired t-test and ANOVA—are designed for different situations (means, not independent proportions).

When you’re comparing two independent groups on a yes/no outcome, you’re looking at two proportions and asking if the difference you observe is real or just due to sampling variation. The best-fitting test is the two-proportion z-test. It evaluates whether the difference between the sample proportions (63% for dog owners vs 53% for cat owners) is large enough to conclude a real difference in the population proportions. It does this by comparing the observed difference to the standard error of that difference under the null hypothesis that the true proportions are equal. Large-sample conditions enable the normal (z) approximation to the binomial, which is why this test is appropriate here. If the counts were very small, Fisher’s exact test or a chi-square test on a 2x2 table might be used instead; the other options—paired t-test and ANOVA—are designed for different situations (means, not independent proportions).