If the sample proportions in Example 4 comparing cancer death rates for aspirin and placebo had sample sizes of only 1000 each, rather than about 11,000 each, then the 95% confidence interval for (p₁ - p₂₎ would be (-0.007, 0.021) rather than (0.003, 0.011). This reflects that
a. When an effect is small, it may take very large samples to have much power for establishing statistical significance.
b. Smaller sample sizes are preferable because there is more of a chance of capturing 0 in a confidence interval.
c. Confidence intervals get wider when sample sizes get larger.
d. The confidence interval based on small sample sizes must be in error because it is impossible for the parameter to take a negative value.