Three types of ecologic variables?
- Aggregate measures (means, proportions of individual characteristics)
- Environmental measures (geographic characteristics)
- Global measures (systemic characteristics)
Synonym for ecological fallacy?
Aggregation bias
2 times when ecological studies are a good choice?
- When we intend to infer at the population level (e.g., for implications and interventions at population level)
- When the variability of the exposure within the population is limited
4 advantages of ecologic studies?
- Greater population coverage (so better generalizability)
- Some exposures better captured
- Greater exposure variation
- Cheap
4 disadvantages of ecologic studies?
- Harder to control confounding
- Secondary data
- Temporal relationship difficult to define
- Ecological fallacy
2 advantages and 1 disadvantage of prospective/concurrent cohort studies?
Advantages:
- Measurement of all variables
- Temporal ordering
Disadvantages:
1. Expensive, in money and time
2 advantages and 1 disadvantage of retrospective/non-concurrent and mixed cohort studies?
Advantages:
- Cheaper
- Linking data registries
Disadvantages:
1. Data already collected, not necessarily of high quality or for research purposes
4 assumptions of case-crossover studies?
- The exposure varies over time
- The exposure has a short induction time and a transient effect
- No cumulative effect of exposure
- Confounding factors do not vary over time
Case-based case control study:
- Sampling method?
- Measure of association?
- At the end of the follow-up
- If rare disease, the OR estimates the risk ratio
Nested case-control study:
- Sampling method?
- Measure of association?
- Concurrently (cumulative)
- Rate ratio (when matched on time analysis)
Case-cohort study:
- Sampling method?
- Measure of association?
- Beginning of follow-up
- Risk ratio (if censoring unrelated to exposure)
- We can using this control group for multiple outcomes
3 assumptions of survival analysis:
- Censoring independent of survival
- No secular trends
- Uniform losses and events
Prevalence odds formula?
Incidence Rate x 1/RR
