Class Notes (1,100,000)
US (440,000)
UC-Irvine (10,000)
PUBHLTH (600)
Lecture 16

PUBHLTH 101 Lecture Notes - Lecture 16: Recall Bias, Observational Error, Sampling Error


Department
Public Health
Course Code
PUBHLTH 101
Professor
David Timberlake
Lecture
16

This preview shows half of the first page. to view the full 2 pages of the document.
LEC 16
SPURIOUS ASSOCATIONS
1. Random error/chance: An alpha level of 0.05 leaves a 1/20 chance of a false-positive
result
2. Systematic (non-random) error: Bias which can occur in a study design, conducting or
analysis of an epidemiologic study
Ex. Recall bias in data collection
CAUSAL ASSOCIATIONS:
Casual associations are REAL associations which are causal
In epidemiology, it is difficult to prove a casual association, particularly for
observational studies
THE WEB OF CAUSATION
Not single cause
Causes of disease are interacting
Illustrates the interconnection
CONSISTENCY
Associations are more likely to be causal if they are observed repeatedly by different
persons, in different places , circumstances and time.
Ex. Meta-analysis of studies of lung cancer of never-smoking females exposed to
spousal smoking, by continent
To show a valid statistical association, we need to assess:
Random error: deviation of results and inferences from the truth, occurring only as a
result of chance
1. Measurement error: due to inexact measuring instruments, subjective measures,
must be random
2. Sampling error: can be assessed w/ confidence intervals, can be reduced w/ a larger
sample size
Systematic error (bias): process at any state of inference
1. Underestimate/overestimate the true effect
2. Produce a spurious association
3. Distort apparent direction of an association
You're Reading a Preview

Unlock to view full version