Review Questions for Chapter 12
�12.1 Introduction
- Name the two types of error
in epidemiologic research.
- How do parameters differ
from estimators?
- Describe the notational
convention used by the text to distinguish parameters from estimators.
- Provide a synonym for systematic
error.
- Provide a synonym for random
error.
- Provide an antonym for biased.
- Provide an antonym for precise.
- List ways in which random
error differs from systematic error.
�12.2 Random Error
- [True or false?] Probability
models are commonly used to adjust for bias in epidemiologic
research.
- [T/F?] Objective and
subjective views of probability are not compatible with each
other.
- What two statistical
procedures are used to address random error in epidemiologic studies?
- Suppose the sample size an
observation study could be expanded to infinitely large. What type of
error would be eliminated. What type of error would not be
eliminated?
- [T/F?] Random error is
expressed as the imprecision in an estimate.
�12.3 Systematic Error
- Name three categories of bias
in epi studies.
- [True or false?]
Nondifferential misclassification bias measures of effect away from the
null.
- [True or false?] A bias
away from null tends to underestimate risks.
- Define confounding.
- List the properties of a
confounder.
- What does the Latin word confundere
mean?
- Use of hospitalized controls
in case-control studies could result in this type of bias.
- Will a large study have less
bias than a small study?
- Subjects who experience an
adverse outcome (cases) tend to give responses about potential causes of
the adverse outcome that differ from those given by those that did not
experience the outcome (non-cases). What type of bias can this cause?
- What type of bias would occur
if the code book for the data from an epidemiologic study was mixed up so
that all exposed individuals were mistakenly identified as non-exposed,
and vice-versa?
- In what situations will an
extraneous risk factor not confound the association between an exposure
and disease?
- This question is from
Rothman, 2002, p. 112. Those who favor representative studies claim that
one cannot generalize a study to a population whose characteristics differ
from those in the study population. Thus, 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. (Clue: If a study
of smoking and lung cancer were conducted in London,
would results apply to those who lived in Paris?)
- In pharmacoepidemiologic
studies, the term "confounding by indication" occurs when those
who are given a certain drug have a different medical condition or
severity of medical condition from those who did not take the drug. Is
this truly a problem related to confounding, or is it better to classify
this as a type of selection bias?
Key