In a two-sample t-test comparing mean blood pressure between a drug group and a placebo group, the 95% confidence interval for mu_drug - mu_placebo is (-12.4, -2.6). Which of the following is a valid conclusion?
- AThe drug has no effect on blood pressure because the interval is negative
- BThe drug significantly reduces blood pressure compared to the placebo because the entire interval is negative and does not contain zeroCorrect
- CThe drug increases blood pressure by between 2.6 and 12.4 mmHg
- DMore data are needed because the interval is too narrow
Why B is right
The entire 95% confidence interval (-12.4, -2.6) is negative, indicating that mu_drug - mu_placebo < 0 with 95% confidence. This means the drug group has a lower mean blood pressure than the placebo group by between 2.6 and 12.4 mmHg. Since zero is not in the interval, the difference is statistically significant.
Why the other 3 choices miss
- A
- Incorrect. A negative interval means the drug reduces blood pressure compared to placebo. 'No effect' would require the interval to contain zero.
- C
- Incorrect. The negative values mean the drug decreases blood pressure, not increases it. The drug group's mean is 2.6 to 12.4 mmHg lower than the placebo group's.
- D
- Incorrect. A narrow confidence interval indicates a precise estimate. There is no requirement for additional data; the interval provides clear evidence of a statistically significant reduction.
A sports physiologist measures the VO2 max of 30 male athletes (mean = 55, SD = 6) and 30 female athletes (mean = 48, SD = 5). She constructs separate 95% confidence intervals: Males: (52.76, 57.24), Females: (46.13, 49.87). She notices the intervals do not overlap and concludes the means are significantly different. A statistician warns this method is:
- ACorrect; non-overlapping intervals always indicate a significant difference, so no further test is needed
- BIncorrect; overlapping intervals always indicate no difference between the two means
- CConservative; non-overlap implies significance, but overlap does not rule it out, so a two-sample test is preferredCorrect
- DInvalid; separate confidence intervals for each group can never be used, even informally, to compare two population means
Why C is right
While non-overlapping confidence intervals do imply a significant difference between the means, the converse is not true. Overlapping intervals do not necessarily mean there is no significant difference. A formal two-sample confidence interval or hypothesis test is the proper way to compare two means. Comparing individual CIs is conservative (may miss real differences).
Why the other 3 choices miss
- A
- Incorrect because while the conclusion is correct in this case, the method is not fully reliable: overlapping intervals can still correspond to a significant difference, so a formal test is still warranted.
- B
- Incorrect because overlapping intervals do not always indicate no difference; the overlap must be substantial relative to the margin of error.
- D
- Incorrect because separate CIs can provide some informal comparison, but the method has limitations and should not replace formal tests.