Lecture 3 (Ecologic Studies) Flashcards
1
Q
Ecologic study
A
- Studies association between exposure and disease where units of analysis are groups (vs individuals)
- Involves assessment of correlation
- Descriptive OR analytic
2
Q
Ecological variables: def and 3 types
A
- Properties of grps, places, or organizations
1) Aggregate variables
2) Environmental variables
3) Global variables
3
Q
Aggregate variables
A
-Summary of observations of individuals in grp (mean, median, proportion, rate)
4
Q
Environmental variables
A
- Physical characteristics of a place
- Usually have analogues at individual level, which vary
- Ex: pollution, rainfall
5
Q
Global variables
A
- Attribute of a grp, organization, or place for which there is no distinct analogue at the individual level
- Ex: pop density, existence of law, etc.
6
Q
Pros: ecological studies
A
- Convenient, inexpensive (2ndary data sources)
- Reflects group averages (substantial variability at individual level)
- Allows analysis of exposure’s effect when it doesn’t vary enough at individual level
- Most appropriate level of analysis for certain variables
7
Q
Linear regression analysis
A
-Finds mathematical model to describe disease variable (y) as a function of exposure variables (x)
y=mx+b or multiple linear regression formula
8
Q
Correlation coefficient
A
- Measure of association indicating degree to which 2 variables have linear relationship
- 1 < r < +1
- Higher absolute value: stronger linear relationship
- Close to 0: little/no linear relationship between variables
- Used for aggregate and individual level variables
9
Q
Limitations of correlation
A
- Doesn’t measure strength of non-linear rel
- n<50 pairs: reliability unclear
- Outliers can skew results
- Correlation =/= causation
10
Q
Coefficient of determination
A
- r2
- Proportion of variance in y explained by x in model
- How good line is at explaining data
- Does not measure strength of slope
11
Q
Ecologic Fallacy/Bias
A
- Association between variables at ecologic level might not translate to individual level
- Prob: heterogeneity of exposure w/in grps (are those exposed the ones developing disease?)
- Major limitation for making causal inferences