Suppose a case-control study could be expanded to be infinitely large. Which sources of error would be eliminated by such a study, and which would not? Suppose that a randomized trial could be infinitely large. Which sources of error would remain in such a trial?
Will a larger study have less bias than a smaller study? Why or why not?
When recall bias occurs, patients who have been afflicted with a medical problem, such as a heart attack, give responses about possible causes of that problem that differ from those given by nonafflicted subjects. Whose responses are thought to be more accurate?
Suppose that in analyzing the data from an epidemiologic study, a computer coding error led to the exposed group being classified as unexposed and the unexposed group being classified as exposed. What specific effect would this error have on the reported results? Is this a bias? If so, what type? If not, what type of error is it?
Explain the difference between a confounding factor and a potential confounding factor. In what situations might a potential confounding factor not end up being a confounding factor?
The incidence rate of cardiovascular disease increases with increasing age. Does that mean that age always confounds studies of cardiovascular disease in the same direction? Why or why not?
The effectiveness of randomization in controlling confounding depends of nutritional education of schoolchildren on their serum cholesterol levels. Suppose that the study involved randomly assigning 10 classrooms with 30 children each to receive a new curriculum and assigning another 10 classrooms with 30 children each to receive the old curriculum. Should this be considered a study that compares two groups with 300 in each group or 10 in each group from the viewpoint of the effectiveness of controlling confounding by randomization?
Confounding by indication arises because those who take a given drug differ for medical reasons from those who do not take the drug. Is this problem truly confounding, or is it more appropriately described as a selection bias?
Those who favor representative studies claim that one should not generalize a study to a population whose characteristics differ from those of the study population. A study of smoking and lung cancer in men would tell nothing about the relation between smoking and lung cancer in women. Give the counterarguments. (Hint: if the study were conducted in London, would the results apply to those who lived in Paris?