Question

10. Match each description to the correct confidence interval in the diagram below. Must get all matches correct. No partial credit. Fail to reject H. P> 0.05. Not statistically significant despite large sample size, which justifies acceptance of the null hypothesis (Cl captures 1 because the RR estimate is close to 1). Fall to reject H. P> 0.05. Although the RR estimate is close to 1, we cannot accept the null hypothesis since the Ci is wide (imprecise). Cannot distinguish benefit from harm. Reject Ho and accept Ha. P<0.05. Not clinically significant. RR estimate is statistically significant but the Ci is wide (imprecise). The risk reduction may be large and real, but the upper limit of the Cl is too close to 1 to take action Fail to reject H. P> 0.05. Not statistically significant due to small sample size. This signifies a probable type II error (CI captures 1 not because the RR estimate is close to 1 but because the interval is wide) Reject Ho and accept HA. P<0.05. Clinically significant, large sample estimate, upper confidence limit far from null hypothesis RR, a blockbuster drug. RR = 1 А B E Benefit Harm

Answer 1

Result 1: This result applies to B. Here RR estimate is not statistically significant (since the CI contains the null value RR = 1 and the interval is narrow.  The size of the box suggests a sufficiently large sample.

Result 2: This result applies to E. Here RR estimate is not statistically significant (since the CI, which is wide and hence, imprecise contains the null value RR = 1 but the null value RR= 1, lies at the centre of the CI i.e the tails extend in equal length on both sides and hence cannot distinguish benefit from harm.

Result 3: This result applies to A. Here RR estimate is statistically significant (since the CI does not contain the null value RR = 1). But we must take note that the CI is wide and the test is not that precise; since it suggests a wide range of probable values. Looking at the size of the plot, we may say that the sample estimate is large enough. Also, although the outcome is beneficial, the upper limit of CI is close to 1.

Result 4: This result applies to C. Here RR estimate is not statistically significant (since the CI contains the null value RR = 1 but we must take note that this is because the CI is wide. The major part of the interval lies on the beneficial side of the outcome and only a small part crossing the null line would result in acceptance of a false null hypothesis.This would result in type II error.

Result 5: This result applies to D. Here RR estimate is clinically significant (since, the outcome completely lies on the beneficial end).We must take note that the CI is not that wide and hence, the test is more precise as compared to A. precise. Looking at the size of the plot, we may say that the sample estimate is large enough. Also, we find that the upper limit of CI is very far from RR = 1.