Statistical Reporting and Interpretation
1. Two researchers conduct separate randomized clinical trials comparing the same new therapy and publish their results in consecutive New England Journal of Medicine papers. The first study found a difference in favor of the new therapy and reported p=.02 to support their contention that this difference was real. The second study used twice as many patients as the first and also observed a difference in favor of the experimental therapy, but reported that p=.21 and therefore that observed difference was not statistically significant. Which statement below is not a possible explanation of the different results?
a. One of the reported p-values must be incorrect, since larger sample sizes generate smaller p-values.
b. The observed difference favoring the new therapy was smaller in the second clinical trial than in the first.
c. Because of differences in designs of the two clinical trials, measurement error and biological variability were greater in the second trial than in the first.
d. The first trial included an adjustment for confounding variables, while the second examined the crude difference between therapies.
e. The first study was not blinded, and diagnostic suspicion bias contaminated the results.
2. An epidemiologic report notes that Reye's syndrome is associated with aspirin usage during a recent febrile illness (p=.034). The association is
a. statistically significant at both a=.05 and a=.01.
b. statistically significant at a=.05 but not at a=.01.
c. statistically significant at a=.03 but not at a=.01.
d. statistically significant at a=.01 but not at a=.05.
e. statistically significant at neither a=.05 nor a=.01.
3. Which one of the following would not be expected to increase the statistical power of a clinical trial to test a new medical treatment?
a. treating a more clinically homogeneous group of patients
b. lengthening the trial to allow enrollment of more patients
c. adopting a more stringent standard of evidence to prove the new treatment works
d. adopting stronger measures to ensure compliance with therapy
e. testing only a subgroup of patients for whom the new drug would
be expected to have the greatest advantage
4. Suppose researchers investigating the relationship between suicide and use of a particular antidepressant medication using a case-control study report an observed odds-ratio of 1.8 and a 90% confidence interval for the true odds-ratio as 0.8-4.05. The appropriate interpretation is that
a. use of the antidepressant was more common among the suicides than among controls, and the association is statistically significant at a=.10.
b. use of the antidepressant was more common among the suicides than among controls, and the association is statistically significant at a=.05.
c. use of the antidepressant was more common among the suicides than among controls, but the association is not statistically significant at a=.10.
d. we can be 90% confident that the suicide rate among users of the antidepressant medication is 80% to 405% higher than among non-users.
e. the statistical power of the case-control study was less than 90%.
5. Two independent studies report estimates and 95% confidence intervals for the same relative risk as 1.0 (CI 0.5-2.0) and 1.5 (CI 1.2-2.0). Which is the best interpretation?
a. The results are contradictory because one
study gives a statistically significant relative risk while the other does not.
b. The results are consistent because both confidence intervals exclude zero.
c. The results are contradictory because one interval is compatible only with an elevated risk, while the other is compatible with either an elevated or a decreased risk.
d. The results are contradictory because no increase in risk was observed in one study while a 50% increase in risk was observed in the other.
e. The results are consistent because results of both studies are compatible with an increase of from 20% to 100% in risk.
6. The statement "p=.05" and the 95% confidence interval 0.0-4.2, each describing the difference in results of two clinical treatments, each summarize results of a collection of hypothesis tests. Precisely describe the collection of hypothesis test results summarized by the p-value and why knowledge of the whole collection of results is useful, and similarly describe the collection of hypothesis test results summarized by the confidence interval and how knowledge of those results is useful.