Question 1

The data from the studies in the figure come from randomised trials of drug treatments.Which RR estimate(s) is/are most likely to be clinically significant? (Select
All that apply)

A)Study 1
B)Study 2
C)Study 3
D)Study 4
E)Study 5

Study 3
Study 4

Accepted Answer

Studies 3 and 4 both show substantial relative effects of the
treatments, while the others RRs are all quite close to 1.0 or the no effect line.However
both of these estimates have quite wide confidence limits, so there is a fair degree of
imprecision in the estimated RRs; and the result from study 3 is not statistically
significant.Thus both should be interpreted cautiously.

Question 2

A large study sample size is important because it:&#10;A)Reduces selection bias in the study&#10;B)Increases study precision&#10;C)Reduces confounding in the study&#10;D)Increases study accuracy&#10;E)None of the above

Accepted Answer

Increases study precision.Larger sample sizes do not influence selection bias
(A), and are also irrelevant to controlling confounding (C).The accuracy of a measure
(D; how close to the 'truth' it comes) is also not affected by increasing sample size.

Question 3

A study found that the relative risk (RR) for the association between air pollution and asthma was RR=1.5, 95%CI (1.2-1.9) with a P value of 0.001.The likelihood that this
Association occurred by chance is:

A)1 in 20
B)1 in 100
C)1 in 1000
D)None of the above

1 in 1000

Accepted Answer

The p‐value tells us the probability that we would see a RR of 1.5 or more if
there were really no association is only about 1 in 1000 times.If it were 1 in 20 (A) the
p‐value would be 0.05; if it were 1 in 100 the p‐value would be 0.01. 11ebbba8_3418_0093_bcf2_b75ae6d62409_TB8436_00

Question 4

Which of the following relative risks (and 95% confidence intervals) is the most precise?&#10;A)1.5 (1.2-1.9)&#10;B)2.1 (0.9-5.3)&#10;C)1.1 (1.0-1.2)&#10;D)0.6 (0.3-0.9)

Accepted Answer

1.1 (1.0-1.2).
The narrower the confidence interval around a point estimate,
the more certain we can be of the true value of whatever it is that we are trying to
measure.A narrow confidence interval suggests a large study where sampling variability
is minimised.

Question 5

You wish to conduct a case&#8208;control study investigating the association between chewing betel quid (from the areca nut) and oral cancer in Kerala.You estimate that&#10;The prevalence of betel quid use in the general population is 20% and based on&#10;Previous studies you expect an odds ratio (OR) for oral cancer of OR=2.0 associated&#10;With betel quid chewing compared to no betel quid chewing.Using the table below,&#10;How large a sample (cases and controls) would you want for this study?&#10;Table: Number of cases required for a case&#8208;control study to detect a statistically&#10;Significant association with varying levels of exposure prevalence1     Number of cases required (assuming 1 control per case), to achieve 80% power at   =&#10;0)05.&#10;A)169&#10;B)338&#10;C)513&#10;D)1026

Accepted Answer

The total sample size required is 338 (169 cases and 169 controls); the prime
assumption here is that your controls give a good representation of the source
population (ideally a random sample from that population), so their exposure
prevalence is close to 20%.

Question 6

Which relative risk (RR) estimate(s) in the figure above is/are statistically significant? (Select all that apply)&#10;A)Study 1&#10;B)Study 2&#10;C)Study 3&#10;D)Study 4&#10;E)Study 5

Accepted Answer

From the figure we can see that studies 2 and 4 are the only ones
where both the lower and upper bounds of the confidence interval are above the null
value (i.e.the confidence interval does not include the null).

Question 7

A study that is able to detect a relative risk of 2.0 between an exposure and an outcome 4 times out of 5 has:&#10;A)Power of 80%&#10;B)An alpha level of 0.05&#10;C)A high probability of Type I error&#10;D)A high probability of Type II error&#10;E)None of the above

Accepted Answer

The power of a study tells us how likely it is that a study will detect an
association if it truly exists, in this case it would detect it 4 out 5 times = 80% of the time.
The probability of making a Type I error (C) or alpha level (B) tells us how likely it is that
we would see an association of a given magnitude if there were really no association.If
the power is 80%, the probability of making a type II error (wrongly concluding there is
no association) is only 20% which is not very high (D).

Question 8

Which RR estimate could be considered both statistically and clinically significant?&#10;A)Study 1&#10;B)Study 2&#10;C)Study 3&#10;D)Study 4&#10;E)Study 5

Accepted Answer

Study 4 is the only one that suggests a large effect and the confidence
interval does not include the null value.

Question 9

When we conduct an hypothesis test we:&#10;A)Assess the probability that the null hypothesis 'there is no association between exposure and outcome' is true&#10;B)Assess the probability that the null hypothesis 'there is no association between exposure and outcome' is false&#10;C)Assess the probability that the alternative hypothesis 'there is an association between exposure and outcome' is true&#10;D)Assess the probability that the alternative hypothesis 'there is an association between exposure and outcome' is false

Accepted Answer

When we conduct an hypothesis test, we assess the probability that we
would have seen a result as (or more) extreme as the observed value if there were
really no association, i.e.if the null hypothesis were true.If this probability is very small
then we can reject the null hypothesis in favour of the alternative hypothesis that there
is an association.

Question 10

Which of the following statements best describes the advantages of confidence intervals over p&#8208;values?&#10;A)Confidence intervals provide an estimate of the likelihood that the association occurred by chance&#10;B)Confidence intervals provide information about the precision of the measure of association&#10;C)Confidence intervals provide an estimate of statistical significance&#10;D)All of the above&#10;E)None of the above

Accepted Answer

Confidence intervals provide information about the precision of the measure
of association.Statements A and C are true of both confidence intervals and p‐values.

Deck 6: Heads or Tails: the Role of Chance